apex-dev 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] (APEXMALHAR-2085) Implement Windowed Operators
Date Mon, 27 Jun 2016 21:19:17 GMT

    [ https://issues.apache.org/jira/browse/APEXMALHAR-2085?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15351836#comment-15351836
] 

ASF GitHub Bot commented on APEXMALHAR-2085:
--------------------------------------------

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

    https://github.com/apache/apex-malhar/pull/319#discussion_r68658247
  
    --- Diff: library/src/main/java/org/apache/apex/malhar/lib/window/impl/AbstractWindowedOperator.java
---
    @@ -0,0 +1,490 @@
    +/**
    + * 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.apex.malhar.lib.window.impl;
    +
    +import java.util.ArrayList;
    +import java.util.Iterator;
    +import java.util.List;
    +import java.util.Map;
    +
    +import javax.validation.ValidationException;
    +
    +import org.joda.time.Duration;
    +import org.slf4j.Logger;
    +import org.slf4j.LoggerFactory;
    +
    +import org.apache.apex.malhar.lib.window.Accumulation;
    +import org.apache.apex.malhar.lib.window.ControlTuple;
    +import org.apache.apex.malhar.lib.window.TriggerOption;
    +import org.apache.apex.malhar.lib.window.Tuple;
    +import org.apache.apex.malhar.lib.window.Window;
    +import org.apache.apex.malhar.lib.window.WindowOption;
    +import org.apache.apex.malhar.lib.window.WindowState;
    +import org.apache.apex.malhar.lib.window.WindowedOperator;
    +import org.apache.apex.malhar.lib.window.WindowedStorage;
    +import org.apache.hadoop.classification.InterfaceStability;
    +
    +import com.google.common.base.Function;
    +
    +import com.datatorrent.api.Context;
    +import com.datatorrent.api.DefaultInputPort;
    +import com.datatorrent.api.DefaultOutputPort;
    +import com.datatorrent.api.annotation.InputPortFieldAnnotation;
    +import com.datatorrent.common.util.BaseOperator;
    +
    +/**
    + * This is the abstract windowed operator class that implements most of the windowing,
triggering, and accumulating
    + * concepts. The subclass of this abstract class is supposed to provide the implementation
of how the accumulated
    + * values are stored in the storage.
    + *
    + * @param <InputT> The type of the input tuple
    + * @param <OutputT> The type of the output tuple
    + * @param <DataStorageT> The type of the data storage
    + * @param <AccumulationT> The type of the accumulation
    + */
    +@InterfaceStability.Evolving
    +public abstract class AbstractWindowedOperator<InputT, OutputT, DataStorageT extends
WindowedStorage, AccumulationT extends Accumulation>
    +    extends BaseOperator implements WindowedOperator<InputT>
    +{
    +
    +  protected WindowOption windowOption;
    +  protected TriggerOption triggerOption;
    +  protected long allowedLatenessMillis = -1;
    +  protected WindowedStorage<WindowState> windowStateMap;
    +
    +  private Function<InputT, Long> timestampExtractor;
    +
    +  private long currentWatermark;
    +  private boolean triggerAtWatermark;
    +  private long earlyTriggerCount;
    +  private long earlyTriggerMillis;
    +  private long lateTriggerCount;
    +  private long lateTriggerMillis;
    +  private long currentDerivedTimestamp = -1;
    +  private long windowWidthMillis;
    +  protected DataStorageT dataStorage;
    +  protected DataStorageT retractionStorage;
    +  protected AccumulationT accumulation;
    +
    +  private static final transient Logger LOG = LoggerFactory.getLogger(AbstractWindowedOperator.class);
    +
    +  public final transient DefaultInputPort<Tuple<InputT>> input = new DefaultInputPort<Tuple<InputT>>()
    +  {
    +    @Override
    +    public void process(Tuple<InputT> tuple)
    +    {
    +      processTuple(tuple);
    +    }
    +  };
    +
    +  // TODO: This port should be removed when Apex Core has native support for custom control
tuples
    +  @InputPortFieldAnnotation(optional = true)
    +  public final transient DefaultInputPort<ControlTuple> controlInput = new DefaultInputPort<ControlTuple>()
    +  {
    +    @Override
    +    public void process(ControlTuple tuple)
    +    {
    +      if (tuple instanceof ControlTuple.Watermark) {
    +        processWatermark((ControlTuple.Watermark)tuple);
    +      }
    +    }
    +  };
    +
    +
    +  // TODO: multiple input ports for join operations
    +
    +  public final transient DefaultOutputPort<Tuple<OutputT>> output = new DefaultOutputPort<>();
    +
    +  // TODO: This port should be removed when Apex Core has native support for custom control
tuples
    +  public final transient DefaultOutputPort<ControlTuple> controlOutput = new DefaultOutputPort<>();
    +
    +  /**
    +   * Process the incoming data tuple
    +   *
    +   * @param tuple
    +   */
    +  public void processTuple(Tuple<InputT> tuple)
    +  {
    +    long timestamp = extractTimestamp(tuple);
    +    if (isTooLate(timestamp)) {
    +      dropTuple(tuple);
    +    } else {
    +      Tuple.WindowedTuple<InputT> windowedTuple = getWindowedValue(tuple);
    +      // do the accumulation
    +      accumulateTuple(windowedTuple);
    +
    +      for (Window window : windowedTuple.getWindows()) {
    +        WindowState windowState = windowStateMap.get(window);
    +        windowState.tupleCount++;
    +        // process any count based triggers
    +        if (windowState.watermarkArrivalTime == -1) {
    +          // watermark has not arrived yet, check for early count based trigger
    +          if (earlyTriggerCount > 0 && (windowState.tupleCount % earlyTriggerCount)
== 0) {
    +            fireTrigger(window, windowState);
    +          }
    +        } else {
    +          // watermark has arrived, check for late count based trigger
    +          if (lateTriggerCount > 0 && (windowState.tupleCount % lateTriggerCount)
== 0) {
    +            fireTrigger(window, windowState);
    +          }
    +        }
    +      }
    +    }
    +  }
    +
    +  @Override
    +  public void setWindowOption(WindowOption windowOption)
    +  {
    +    this.windowOption = windowOption;
    +    if (this.windowOption instanceof WindowOption.GlobalWindow) {
    +      windowStateMap.put(Window.GLOBAL_WINDOW, new WindowState());
    +    }
    +  }
    +
    +  @Override
    +  public void setTriggerOption(TriggerOption triggerOption)
    +  {
    +    this.triggerOption = triggerOption;
    +    for (TriggerOption.Trigger trigger : triggerOption.getTriggerList()) {
    +      switch (trigger.getWatermarkOpt()) {
    +        case ON_TIME:
    +          triggerAtWatermark = true;
    +          break;
    +        case EARLY:
    +          if (trigger instanceof TriggerOption.TimeTrigger) {
    +            earlyTriggerMillis = ((TriggerOption.TimeTrigger)trigger).getDuration().getMillis();
    +          } else if (trigger instanceof TriggerOption.CountTrigger) {
    +            earlyTriggerCount = ((TriggerOption.CountTrigger)trigger).getCount();
    +          }
    +          break;
    +        case LATE:
    +          if (trigger instanceof TriggerOption.TimeTrigger) {
    +            lateTriggerMillis = ((TriggerOption.TimeTrigger)trigger).getDuration().getMillis();
    +          } else if (trigger instanceof TriggerOption.CountTrigger) {
    +            lateTriggerCount = ((TriggerOption.CountTrigger)trigger).getCount();
    +          }
    +          break;
    +        default:
    +          throw new RuntimeException("Unexpected watermark option: " + trigger.getWatermarkOpt());
    +      }
    +    }
    +  }
    +
    +  @Override
    +  public void setAllowedLateness(Duration allowedLateness)
    +  {
    +    this.allowedLatenessMillis = allowedLateness.getMillis();
    +  }
    +
    +  /**
    +   * This method sets the storage for the data for each window
    +   *
    +   * @param storageAgent
    +   */
    +  public void setDataStorage(DataStorageT storageAgent)
    +  {
    +    this.dataStorage = storageAgent;
    +  }
    +
    +  /**
    +   * This method sets the storage for the retraction data for each window. Only used
when the accumulation mode is ACCUMULATING_AND_RETRACTING
    +   *
    +   * @param storageAgent
    +   */
    +  public void setRetractionStorage(DataStorageT storageAgent)
    +  {
    +    this.retractionStorage = storageAgent;
    +  }
    +
    +  /**
    +   * Sets the accumulation, which basically tells the WindowedOperator what to do if
a new tuple comes in and what
    +   * to put in the pane when a trigger is fired
    +   *
    +   * @param accumulation
    +   */
    +  public void setAccumulation(AccumulationT accumulation)
    +  {
    +    this.accumulation = accumulation;
    +  }
    +
    +  @Override
    +  public void setWindowStateStorage(WindowedStorage<WindowState> storageAgent)
    +  {
    +    this.windowStateMap = storageAgent;
    +  }
    +
    +  @Override
    +  public void setTimestampExtractor(Function<InputT, Long> timestampExtractor)
    +  {
    +    this.timestampExtractor = timestampExtractor;
    +  }
    +
    +  public void validate() throws ValidationException
    +  {
    +    if (accumulation == null) {
    +      throw new ValidationException("Accumulation must be set");
    +    }
    +    if (dataStorage == null) {
    +      throw new ValidationException("Data storage must be set");
    +    }
    +    if (windowStateMap == null) {
    +      throw new ValidationException("Window state storage must be set");
    +    }
    +    if (triggerOption != null) {
    +      if (triggerOption.isFiringOnlyUpdatedPanes()) {
    +        if (retractionStorage == null) {
    +          throw new ValidationException("A retraction storage is required for firingOnlyUpdatedPanes
option");
    +        }
    +        if (triggerOption.getAccumulationMode() == TriggerOption.AccumulationMode.DISCARDING)
{
    +          throw new ValidationException("DISCARDING accumulation mode is not valid for
firingOnlyUpdatedPanes option");
    +        }
    +      }
    +      if (triggerOption.getAccumulationMode() == TriggerOption.AccumulationMode.ACCUMULATING_AND_RETRACTING
&&
    +          retractionStorage == null) {
    +        throw new ValidationException("A retraction storage is required for ACCUMULATING_AND_RETRACTING
accumulation mode");
    +      }
    +    }
    +  }
    +
    +  @Override
    +  public Tuple.WindowedTuple<InputT> getWindowedValue(Tuple<InputT> input)
    +  {
    +    Tuple.WindowedTuple<InputT> windowedTuple = new Tuple.WindowedTuple<>();
    +    windowedTuple.setValue(input.getValue());
    +    windowedTuple.setTimestamp(extractTimestamp(input));
    +    assignWindows(windowedTuple.getWindows(), input);
    +    return windowedTuple;
    +  }
    +
    +  private long extractTimestamp(Tuple<InputT> tuple)
    +  {
    +    if (timestampExtractor == null) {
    +      if (tuple instanceof Tuple.TimestampedTuple) {
    +        return ((Tuple.TimestampedTuple)tuple).getTimestamp();
    +      } else {
    +        return 0;
    +      }
    +    } else {
    +      return timestampExtractor.apply(tuple.getValue());
    +    }
    +  }
    +
    +  private void assignWindows(List<Window> windows, Tuple<InputT> inputTuple)
    +  {
    +    if (windowOption instanceof WindowOption.GlobalWindow) {
    +      windows.add(Window.GLOBAL_WINDOW);
    +    } else {
    +      long timestamp = extractTimestamp(inputTuple);
    +      if (windowOption instanceof WindowOption.TimeWindows) {
    +
    +        for (Window.TimeWindow window : getTimeWindowsForTimestamp(timestamp)) {
    +          if (!windowStateMap.containsWindow(window)) {
    +            windowStateMap.put(window, new WindowState());
    +          }
    +          windows.add(window);
    +        }
    +      } else if (windowOption instanceof WindowOption.SessionWindows) {
    +        assignSessionWindows(windows, timestamp, inputTuple);
    +      }
    +    }
    +  }
    +
    +  protected void assignSessionWindows(List<Window> windows, long timestamp, Tuple<InputT>
inputTuple)
    +  {
    +    throw new UnsupportedOperationException();
    +  }
    +
    +  /**
    +   * Returns the list of windows TimeWindows for the given timestamp.
    +   * If we are doing sliding windows, this will return multiple windows. Otherwise, only
one window will be returned.
    +   * Note that this method does not apply to SessionWindows.
    +   *
    +   * @param timestamp
    +   * @return
    +   */
    +  private List<Window.TimeWindow> getTimeWindowsForTimestamp(long timestamp)
    +  {
    +    List<Window.TimeWindow> windows = new ArrayList<>();
    +    if (windowOption instanceof WindowOption.TimeWindows) {
    +      long durationMillis = ((WindowOption.TimeWindows)windowOption).getDuration().getMillis();
    +      long beginTimestamp = timestamp - timestamp % durationMillis;
    +      windows.add(new Window.TimeWindow(beginTimestamp, durationMillis));
    +      if (windowOption instanceof WindowOption.SlidingTimeWindows) {
    +        long slideBy = ((WindowOption.SlidingTimeWindows)windowOption).getSlideByDuration().getMillis();
    +        // add the sliding windows front and back
    +        // Note: this messes up the order of the window and we might want to revisit
this if the order of the windows
    +        // matter
    +        for (long slideBeginTimestamp = beginTimestamp - slideBy;
    +            slideBeginTimestamp <= timestamp && timestamp < slideBeginTimestamp
+ durationMillis;
    +            slideBeginTimestamp -= slideBy) {
    +          windows.add(new Window.TimeWindow(slideBeginTimestamp, durationMillis));
    +        }
    +        for (long slideBeginTimestamp = beginTimestamp + slideBy;
    +            slideBeginTimestamp <= timestamp && timestamp < slideBeginTimestamp
+ durationMillis;
    +            slideBeginTimestamp += slideBy) {
    +          windows.add(new Window.TimeWindow(slideBeginTimestamp, durationMillis));
    +        }
    +      }
    +    } else {
    +      throw new IllegalStateException("Unexpected WindowOption");
    +    }
    +    return windows;
    +  }
    +
    +  @Override
    +  public boolean isTooLate(long timestamp)
    +  {
    +    return allowedLatenessMillis < 0 ? false : (timestamp < currentWatermark -
allowedLatenessMillis);
    +  }
    +
    +  @Override
    +  public void dropTuple(Tuple<InputT> input)
    +  {
    +    // do nothing
    +    LOG.debug("Dropping late tuple {}", input);
    +  }
    +
    +
    +  @Override
    +  public void processWatermark(ControlTuple.Watermark watermark)
    +  {
    +    currentWatermark = watermark.getTimestamp();
    +    long horizon = currentWatermark - allowedLatenessMillis;
    +    if (allowedLatenessMillis >= 0) {
    +      // purge window that are too late to accept any more input
    +      dataStorage.removeUpTo(horizon);
    --- End diff --
    
    Before removing from storage, should we check whether triggers are fired and data is send
to downstream for window that is too late?


> Implement Windowed Operators
> ----------------------------
>
>                 Key: APEXMALHAR-2085
>                 URL: https://issues.apache.org/jira/browse/APEXMALHAR-2085
>             Project: Apache Apex Malhar
>          Issue Type: New Feature
>            Reporter: Siyuan Hua
>            Assignee: David Yan
>
> As per our recent several discussions in the community. A group of Windowed Operators
that delivers the window semantic follows the google Data Flow model(https://cloud.google.com/dataflow/)
is very important. 
> The operators should be designed and implemented in a way for 
> High-level API
> Beam translation
> Easy to use with other popular operator
> {panel:title=Operator Hierarchy}
> Hierarchy of the operators,
> The windowed operators should cover all possible transformations that require window,
and batch processing is also considered as special window called global window
> {code}
>                    +-------------------+
>        +---------> |  WindowedOperator | <--------+
>        |           +--------+----------+          |
>        |                    ^      ^--------------------------------+
>        |                    |                     |                 |
>        |                    |                     |                 |
> +------+--------+    +------+------+      +-------+-----+    +------+-----+
> |CombineOperator|    |GroupOperator|      |KeyedOperator|    |JoinOperator|
> +---------------+    +-------------+      +------+------+    +-----+------+
>                                    +---------^   ^                 ^
>                                    |             |                 |
>                           +--------+---+   +-----+----+       +----+----+
>                           |KeyedCombine|   |KeyedGroup|       | CoGroup |
>                           +------------+   +----------+       +---------+
> {code}
> Combine operation includes all operations that combine all tuples in one window into
one or small number of tuples, Group operation group all tuples in one window, Join and CoGroup
are used to join and group tuples from different inputs.
> {panel}
> {panel:title=Components}
> * Window Component
> It includes configuration, window state that should be checkpointed, etc. It should support
NonMergibleWindow(fixed or slide) MergibleWindow(Session)
> * Trigger
> It should support early trigger, late trigger with customizable trigger behaviour 
> * Other related components:
> ** Watermark generator, can be plugged into input source to generate watermark
> ** Tuple schema support:
> It should handle either predefined tuple type or give a declarative API to describe the
user defined tuple class
> {panel}
> Most component API should be reused in High-Level API
> This is the umbrella ticket, separate tickets would be created for different components
and operators respectively 



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

Mime
View raw message