Add Platform test to show Hadoop and Tez partitioning difference

cascading-hadoop/src/test/shared/cascading/flow/hadoop/PartitionNodesTest.java

@@ -0,0 +1,94 @@
+package cascading.flow.hadoop;
+
+import java.util.HashMap;
+import java.util.Map;
+import java.util.concurrent.TimeUnit;
+
+import org.junit.Test;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import cascading.PlatformTestCase;
+import cascading.flow.Flow;
+import cascading.flow.LockingFlowListener;
+import cascading.operation.Function;
+import cascading.operation.Identity;
+import cascading.operation.aggregator.Average;
+import cascading.operation.aggregator.Count;
+import cascading.operation.regex.RegexSplitter;
+import cascading.pipe.Each;
+import cascading.pipe.Every;
+import cascading.pipe.GroupBy;
+import cascading.pipe.Merge;
+import cascading.pipe.Pipe;
+import cascading.scheme.hadoop.TextLine;
+import cascading.tap.SinkMode;
+import cascading.tap.Tap;
+import cascading.tap.hadoop.Hfs;
+import cascading.tuple.Fields;
+import cascading.util.Util;
+
+import static data.InputData.inputFileLower;
+import static data.InputData.inputFileUpper;
+
+/**
+ * To demonstrate PartitionNodes difference between MR and Tez
+ * In Hadoop and Tez this results in a single PartitionStep.
+ * Where the two differ is Hadoop partitions into 2 nodes. First node (map) reads from the two sourceTaps, performs the pipeLeft & pipeRight transforms
+ * Second node (reduce) does the merge, each, groupBy, every, every, each, each
+ * In case of Tez, we have 4 nodes. First two are equivalent to the Hadoop node 1, read from tap + each
+ * Next node just does the merge, each, groupBy. Node after that does the every, every, each, each.
+ * As we don't have the aggregations following the groupBy on the same nodes, we end up streaming a lot of tuples to disk.
+ */
+public class PartitionNodesTest extends PlatformTestCase {
+
+  private static final Logger LOG = LoggerFactory.getLogger( PartitionNodesTest.class );
+
+  public PartitionNodesTest()
+  {
+    super( true ); // must be run in cluster mode
+  }
+
+  @Test
+  public void testGraphDifference() throws Exception
+  {
+    if( !getPlatform().isUseCluster() )
+      return;
+
+    getPlatform().copyFromLocal( inputFileLower );
+    getPlatform().copyFromLocal( inputFileUpper );
+
+    Tap sourceLeft = new Hfs( new TextLine( new Fields( "offset", "line" ) ), inputFileLower );
+    Tap sourceRight = new Hfs( new TextLine( new Fields( "offset", "line" ) ), inputFileUpper );
+    Map sources = new HashMap();
+    sources.put( "left", sourceLeft );
+    sources.put( "right", sourceRight );
+
+    Function splitter = new RegexSplitter( new Fields( "num", "char" ), " " );
+    Pipe pipeLeft = new Each( new Pipe( "left" ), new Fields( "line" ), splitter );
+    Pipe pipeRight = new Each( new Pipe( "right" ), new Fields( "line" ), splitter );
+
+    Pipe merge = new Merge("merge", pipeLeft, pipeRight);
+    Pipe mergeEach = new Each(merge, new Identity());
+    Pipe groupBy = new GroupBy(mergeEach, new Fields("num"));
+    Pipe ev1 = new Every(groupBy, new Average());
+    Pipe ev2 = new Every(ev1, new Count());
+    Pipe evEach = new Each(ev2, new Identity());
+    Pipe evEach2 = new Each(evEach, new Identity());
+
+    Tap sink = new Hfs( new TextLine(), getOutputPath( "result" ), SinkMode.REPLACE );
+
+    final Flow flow = getPlatform().getFlowConnector( getProperties() ).connect( sources, sink, evEach2 );
+
+    final LockingFlowListener listener = new LockingFlowListener();
+
+    flow.addListener( listener );
+
+    LOG.info( "calling start" );
+    flow.start();
+
+    Util.safeSleep( 90000 );
+
+    assertTrue( "did not start", listener.started.tryAcquire( 60, TimeUnit.SECONDS ) );
+  }
+}