org.apache.flink.optimizer.plan.SingleInputPlanNode Java Examples

/**
 * Simple Job: Map -> Reduce -> Map -> Reduce. All functions preserve all fields (hence all properties).
 * 
 * Increases parallelism between 2nd map and 2nd reduce, so the hash partitioning from 1st reduce is not reusable.
 * Expected to re-establish partitioning between map and reduce (hash).
 */
@Test
public void checkPropertyHandlingWithIncreasingGlobalParallelism2() {
	final int p = DEFAULT_PARALLELISM;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);
	DataSet<Long> set1 = env.generateSequence(0,1).setParallelism(p);

	set1.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map1")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p).name("Reduce1")
			.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map2")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Reduce2")
			.output(new DiscardingOutputFormat<Long>()).setParallelism(p * 2).name("Sink");

	Plan plan = env.createProgramPlan();
	
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);
	
	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();
	
	ShipStrategyType mapIn = map2Node.getInput().getShipStrategy();
	ShipStrategyType reduceIn = red2Node.getInput().getShipStrategy();
	
	Assert.assertEquals("Invalid ship strategy for an operator.", ShipStrategyType.FORWARD, mapIn);
	Assert.assertEquals("Invalid ship strategy for an operator.", ShipStrategyType.PARTITION_HASH, reduceIn);
}
 

@Test
public void testPipelineBreakerWithBroadcastVariable() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		env.getConfig().setExecutionMode(ExecutionMode.PIPELINED);
		env.setParallelism(64);
		
		DataSet<Long> source = env.generateSequence(1, 10).map(new IdentityMapper<Long>());
		
		DataSet<Long> result = source.map(new IdentityMapper<Long>())
									.map(new IdentityMapper<Long>())
										.withBroadcastSet(source, "bc");
		
		result.output(new DiscardingOutputFormat<Long>());
		
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode mapper = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode mapperInput = (SingleInputPlanNode) mapper.getInput().getSource();
		
		assertEquals(TempMode.NONE, mapper.getInput().getTempMode());
		assertEquals(TempMode.NONE, mapper.getBroadcastInputs().get(0).getTempMode());
		
		assertEquals(DataExchangeMode.BATCH, mapperInput.getInput().getDataExchangeMode());
		assertEquals(DataExchangeMode.BATCH, mapper.getBroadcastInputs().get(0).getDataExchangeMode());
		
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testNoPartialSolutionFoundSingleInputOnly() {
	try {
		SourcePlanNode target = new SourcePlanNode(getSourceNode(), "Source");
		
		SourcePlanNode otherTarget = new SourcePlanNode(getSourceNode(), "Source");
		
		Channel toMap1 = new Channel(target);
		toMap1.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toMap1.setLocalStrategy(LocalStrategy.NONE);
		SingleInputPlanNode map1 = new SingleInputPlanNode(getMapNode(), "Mapper 1", toMap1, DriverStrategy.MAP);
		
		Channel toMap2 = new Channel(map1);
		toMap2.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toMap2.setLocalStrategy(LocalStrategy.NONE);
		SingleInputPlanNode map2 = new SingleInputPlanNode(getMapNode(), "Mapper 2", toMap2, DriverStrategy.MAP);
		
		{
			GlobalProperties gp = new GlobalProperties();
			LocalProperties lp = new LocalProperties();
			
			FeedbackPropertiesMeetRequirementsReport report = map2.checkPartialSolutionPropertiesMet(otherTarget, gp, lp);
			assertTrue(report == NO_PARTIAL_SOLUTION);
		}
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

private void checkStandardStrategies(SingleInputPlanNode map, DualInputPlanNode join, SingleInputPlanNode combiner,
		SingleInputPlanNode reducer, SinkPlanNode sink) {
	// check ship strategies that are always fix
	Assert.assertEquals(ShipStrategyType.FORWARD, map.getInput().getShipStrategy());
	Assert.assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());

	// check the driver strategies that are always fix
	Assert.assertEquals(DriverStrategy.FLAT_MAP, map.getDriverStrategy());
	Assert.assertEquals(DriverStrategy.SORTED_GROUP_REDUCE, reducer.getDriverStrategy());
	Assert.assertEquals(DriverStrategy.NONE, sink.getDriverStrategy());
	if (combiner != null) {
		Assert.assertEquals(DriverStrategy.SORTED_GROUP_COMBINE, combiner.getDriverStrategy());
		Assert.assertEquals(LocalStrategy.NONE, combiner.getInput().getLocalStrategy());
	}
}
 

@Test
public void checkPropertyHandlingWithDecreasingParallelism() {
	final int p = DEFAULT_PARALLELISM;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);

	env
		.generateSequence(0, 1).setParallelism(p * 2)
		.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Map1")
		.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Reduce1")
		.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map2")
		.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p).name("Reduce2")
		.output(new DiscardingOutputFormat<Long>()).setParallelism(p).name("Sink");

	Plan plan = env.createProgramPlan();
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);

	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();

	Assert.assertTrue("The no sorting local strategy.",
			LocalStrategy.SORT == red2Node.getInput().getLocalStrategy() ||
					LocalStrategy.SORT == map2Node.getInput().getLocalStrategy());

	Assert.assertTrue("The no partitioning ship strategy.",
			ShipStrategyType.PARTITION_HASH == red2Node.getInput().getShipStrategy() ||
					ShipStrategyType.PARTITION_HASH == map2Node.getInput().getShipStrategy());
}
 

@Test
public void testDistinctPreservesPartitioningOfDistinctFields() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		env.setParallelism(4);
		
		@SuppressWarnings("unchecked")
		DataSet<Tuple2<Long, Long>> data = env.fromElements(new Tuple2<Long, Long>(0L, 0L), new Tuple2<Long, Long>(1L, 1L))
				.map(new IdentityMapper<Tuple2<Long,Long>>()).setParallelism(4);
		
		data.distinct(0)
			.groupBy(0)
			.sum(1)
			.output(new DiscardingOutputFormat<Tuple2<Long, Long>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode distinctReducer = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		
		// reducer can be forward, reuses partitioning from distinct
		assertEquals(ShipStrategyType.FORWARD, reducer.getInput().getShipStrategy());
		
		// distinct reducer is partitioned
		assertEquals(ShipStrategyType.PARTITION_HASH, distinctReducer.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

private boolean checkBroadcastShipStrategies(DualInputPlanNode join, SingleInputPlanNode reducer,
		SingleInputPlanNode combiner) {
	if (ShipStrategyType.BROADCAST == join.getInput1().getShipStrategy() &&
		ShipStrategyType.FORWARD == join.getInput2().getShipStrategy() &&
		ShipStrategyType.PARTITION_HASH == reducer.getInput().getShipStrategy()) {

		// check combiner
		Assert.assertNotNull("Plan should have a combiner", combiner);
		Assert.assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
		return true;
	} else {
		return false;
	}
}
 

@Test
public void testCustomPartitioningTupleReduce() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		DataSet<Tuple2<Integer, Integer>> data = env.fromElements(new Tuple2<Integer, Integer>(0, 0))
				.rebalance().setParallelism(4);
		
		data.groupBy(0).withPartitioner(new TestPartitionerInt())
			.reduce(new SelectOneReducer<Tuple2<Integer,Integer>>())
			.output(new DiscardingOutputFormat<Tuple2<Integer, Integer>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void checkPropertyHandlingWithDecreasingParallelism() {
	final int p = DEFAULT_PARALLELISM;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);

	env
		.generateSequence(0, 1).setParallelism(p * 2)
		.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Map1")
		.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Reduce1")
		.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map2")
		.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p).name("Reduce2")
		.output(new DiscardingOutputFormat<Long>()).setParallelism(p).name("Sink");

	Plan plan = env.createProgramPlan();
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);

	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();

	Assert.assertTrue("The no sorting local strategy.",
			LocalStrategy.SORT == red2Node.getInput().getLocalStrategy() ||
					LocalStrategy.SORT == map2Node.getInput().getLocalStrategy());

	Assert.assertTrue("The no partitioning ship strategy.",
			ShipStrategyType.PARTITION_HASH == red2Node.getInput().getShipStrategy() ||
					ShipStrategyType.PARTITION_HASH == map2Node.getInput().getShipStrategy());
}
 

@Test
public void testNoPartialSolutionFoundTwoInputOperator() {
	try {
		SourcePlanNode target = new SourcePlanNode(getSourceNode(), "Partial Solution");

		SourcePlanNode source1 = new SourcePlanNode(getSourceNode(), "Source 1");
		SourcePlanNode source2 = new SourcePlanNode(getSourceNode(), "Source 2");
		
		Channel toMap1 = new Channel(source1);
		toMap1.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toMap1.setLocalStrategy(LocalStrategy.NONE);
		SingleInputPlanNode map1 = new SingleInputPlanNode(getMapNode(), "Mapper 1", toMap1, DriverStrategy.MAP);
		
		Channel toMap2 = new Channel(source2);
		toMap2.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toMap2.setLocalStrategy(LocalStrategy.NONE);
		SingleInputPlanNode map2 = new SingleInputPlanNode(getMapNode(), "Mapper 2", toMap2, DriverStrategy.MAP);
		
		Channel toJoin1 = new Channel(map1);
		Channel toJoin2 = new Channel(map2);
		
		toJoin1.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toJoin1.setLocalStrategy(LocalStrategy.NONE);
		toJoin2.setShipStrategy(ShipStrategyType.FORWARD, DataExchangeMode.PIPELINED);
		toJoin2.setLocalStrategy(LocalStrategy.NONE);
		
		DualInputPlanNode join = new DualInputPlanNode(getJoinNode(), "Join", toJoin1, toJoin2, DriverStrategy.HYBRIDHASH_BUILD_FIRST);
		
		FeedbackPropertiesMeetRequirementsReport report = join.checkPartialSolutionPropertiesMet(target, new GlobalProperties(), new LocalProperties());
		assertEquals(NO_PARTIAL_SOLUTION, report);
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testCustomPartitioningKeySelectorGroupReduceSorted() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		DataSet<Tuple3<Integer, Integer, Integer>> data = env.fromElements(new Tuple3<Integer, Integer, Integer>(0, 0, 0))
				.rebalance().setParallelism(4);
		
		data.groupBy(new TestKeySelector<Tuple3<Integer,Integer,Integer>>())
			.withPartitioner(new TestPartitionerInt())
			.sortGroup(new TestKeySelector<Tuple3<Integer, Integer, Integer>>(), Order.ASCENDING)
			.reduceGroup(new IdentityGroupReducerCombinable<Tuple3<Integer,Integer,Integer>>())
			.output(new DiscardingOutputFormat<Tuple3<Integer, Integer, Integer>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testDistinctPreservesPartitioningOfDistinctFields() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		env.setParallelism(4);
		
		@SuppressWarnings("unchecked")
		DataSet<Tuple2<Long, Long>> data = env.fromElements(new Tuple2<Long, Long>(0L, 0L), new Tuple2<Long, Long>(1L, 1L))
				.map(new IdentityMapper<Tuple2<Long,Long>>()).setParallelism(4);
		
		data.distinct(0)
			.groupBy(0)
			.sum(1)
			.output(new DiscardingOutputFormat<Tuple2<Long, Long>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode distinctReducer = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		
		// reducer can be forward, reuses partitioning from distinct
		assertEquals(ShipStrategyType.FORWARD, reducer.getInput().getShipStrategy());
		
		// distinct reducer is partitioned
		assertEquals(ShipStrategyType.PARTITION_HASH, distinctReducer.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testBreakerForDependentVariable() {
		try {
			ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
			
			DataSet<String> source1 = env.fromElements("test");
			
			source1.map(new IdentityMapper<String>()).map(new IdentityMapper<String>()).withBroadcastSet(source1, "some name")
					.output(new DiscardingOutputFormat<String>());
			
			Plan p = env.createProgramPlan();
			OptimizedPlan op = compileNoStats(p);
			
			SinkPlanNode sink = op.getDataSinks().iterator().next();
			SingleInputPlanNode mapper = (SingleInputPlanNode) sink.getInput().getSource();
			SingleInputPlanNode beforeMapper = (SingleInputPlanNode) mapper.getInput().getSource();

			assertEquals(TempMode.NONE, mapper.getInput().getTempMode());
			assertEquals(TempMode.NONE, beforeMapper.getInput().getTempMode());
			assertEquals(TempMode.NONE, mapper.getBroadcastInputs().get(0).getTempMode());

			assertEquals(DataExchangeMode.PIPELINED, mapper.getInput().getDataExchangeMode());
			assertEquals(DataExchangeMode.BATCH, beforeMapper.getInput().getDataExchangeMode());
			assertEquals(DataExchangeMode.BATCH, mapper.getBroadcastInputs().get(0).getDataExchangeMode());
		}
		catch (Exception e) {
			e.printStackTrace();
			fail(e.getMessage());
		}
}
 

@Test
public void testRangePartitionOperatorPreservesFields() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

		DataSet<Tuple2<Long, Long>> data = env.fromCollection(Collections.singleton(new Tuple2<>(0L, 0L)));

		data.partitionByRange(1)
			.groupBy(1)
			.reduceGroup(new IdentityGroupReducerCombinable<Tuple2<Long,Long>>())
			.output(new DiscardingOutputFormat<Tuple2<Long, Long>>());

		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);

		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode partitionNode = (SingleInputPlanNode)reducer.getInput().getSource();
		SingleInputPlanNode partitionIDRemover = (SingleInputPlanNode) partitionNode.getInput().getSource();

		assertEquals(ShipStrategyType.FORWARD, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, partitionNode.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, partitionIDRemover.getInput().getShipStrategy());

		SourcePlanNode sourcePlanNode = op.getDataSources().iterator().next();
		List<Channel> sourceOutgoingChannels = sourcePlanNode.getOutgoingChannels();
		assertEquals(2, sourceOutgoingChannels.size());
		assertEquals(ShipStrategyType.FORWARD, sourceOutgoingChannels.get(0).getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, sourceOutgoingChannels.get(1).getShipStrategy());
		assertEquals(DataExchangeMode.PIPELINED, sourceOutgoingChannels.get(0).getDataExchangeMode());
		assertEquals(DataExchangeMode.BATCH, sourceOutgoingChannels.get(1).getDataExchangeMode());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testPartitioningReused() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		@SuppressWarnings("unchecked")
		DataSet<Tuple3<Long, Long, Long>> input = env.fromElements(new Tuple3<Long, Long, Long>(0L, 0L, 0L));
		
		input
			.groupBy(0).sum(1)
			.groupBy(0, 1).sum(2)
			.output(new DiscardingOutputFormat<Tuple3<Long, Long, Long>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		
		SingleInputPlanNode agg2Reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode agg1Reducer = (SingleInputPlanNode) agg2Reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, agg2Reducer.getInput().getShipStrategy());
		
		assertEquals(ShipStrategyType.PARTITION_HASH, agg1Reducer.getInput().getShipStrategy());
		assertEquals(new FieldList(0), agg1Reducer.getInput().getShipStrategyKeys());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

/**
 * Simple Job: Map -> Reduce -> Map -> Reduce. All functions preserve all fields (hence all properties).
 * 
 * Increases parallelism between 1st reduce and 2nd map, such that more tasks are on one instance.
 * Expected to re-establish partitioning between map and reduce via a local hash.
 */
@Test
public void checkPropertyHandlingWithIncreasingLocalParallelism() {
	final int p = DEFAULT_PARALLELISM * 2;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);
	DataSet<Long> set1 = env.generateSequence(0,1).setParallelism(p);

	set1.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map1")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p).name("Reduce1")
			.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Map2")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Reduce2")
			.output(new DiscardingOutputFormat<Long>()).setParallelism(p * 2).name("Sink");

	Plan plan = env.createProgramPlan();
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);
	
	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();
	
	ShipStrategyType mapIn = map2Node.getInput().getShipStrategy();
	ShipStrategyType reduceIn = red2Node.getInput().getShipStrategy();
	
	Assert.assertTrue("Invalid ship strategy for an operator.", 
			(ShipStrategyType.PARTITION_RANDOM ==  mapIn && ShipStrategyType.PARTITION_HASH == reduceIn) || 
			(ShipStrategyType.PARTITION_HASH == mapIn && ShipStrategyType.FORWARD == reduceIn));
}
 

@Test
public void testCustomPartitioningTupleGroupReduce() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		DataSet<Tuple2<Integer, Integer>> data = env.fromElements(new Tuple2<Integer, Integer>(0, 0))
				.rebalance().setParallelism(4);
		
		data.groupBy(0).withPartitioner(new TestPartitionerInt())
			.reduceGroup(new IdentityGroupReducerCombinable<Tuple2<Integer,Integer>>())
			.output(new DiscardingOutputFormat<Tuple2<Integer, Integer>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testBreakerForDependentVariable() {
		try {
			ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
			
			DataSet<String> source1 = env.fromElements("test");
			
			source1.map(new IdentityMapper<String>()).map(new IdentityMapper<String>()).withBroadcastSet(source1, "some name")
					.output(new DiscardingOutputFormat<String>());
			
			Plan p = env.createProgramPlan();
			OptimizedPlan op = compileNoStats(p);
			
			SinkPlanNode sink = op.getDataSinks().iterator().next();
			SingleInputPlanNode mapper = (SingleInputPlanNode) sink.getInput().getSource();
			SingleInputPlanNode beforeMapper = (SingleInputPlanNode) mapper.getInput().getSource();

			assertEquals(TempMode.NONE, mapper.getInput().getTempMode());
			assertEquals(TempMode.NONE, beforeMapper.getInput().getTempMode());
			assertEquals(TempMode.NONE, mapper.getBroadcastInputs().get(0).getTempMode());

			assertEquals(DataExchangeMode.PIPELINED, mapper.getInput().getDataExchangeMode());
			assertEquals(DataExchangeMode.BATCH, beforeMapper.getInput().getDataExchangeMode());
			assertEquals(DataExchangeMode.BATCH, mapper.getBroadcastInputs().get(0).getDataExchangeMode());
		}
		catch (Exception e) {
			e.printStackTrace();
			fail(e.getMessage());
		}
}
 

@Test
public void testCustomPartitioningTupleGroupReduce() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		DataSet<Tuple2<Integer, Integer>> data = env.fromElements(new Tuple2<Integer, Integer>(0, 0))
				.rebalance().setParallelism(4);
		
		data.groupBy(0).withPartitioner(new TestPartitionerInt())
			.reduceGroup(new IdentityGroupReducerCombinable<Tuple2<Integer,Integer>>())
			.output(new DiscardingOutputFormat<Tuple2<Integer, Integer>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testCustomPartitioningTupleReduce() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		DataSet<Tuple2<Integer, Integer>> data = env.fromElements(new Tuple2<Integer, Integer>(0, 0))
				.rebalance().setParallelism(4);
		
		data.groupBy(0).withPartitioner(new TestPartitionerInt())
			.reduce(new SelectOneReducer<Tuple2<Integer,Integer>>())
			.output(new DiscardingOutputFormat<Tuple2<Integer, Integer>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.PARTITION_CUSTOM, reducer.getInput().getShipStrategy());
		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

private boolean checkRepartitionShipStrategies(DualInputPlanNode join, SingleInputPlanNode reducer,
		SingleInputPlanNode combiner) {
	if (ShipStrategyType.PARTITION_HASH == join.getInput1().getShipStrategy() &&
		ShipStrategyType.PARTITION_HASH == join.getInput2().getShipStrategy() &&
		ShipStrategyType.FORWARD == reducer.getInput().getShipStrategy()) {

		// check combiner
		Assert.assertNull("Plan should not have a combiner", combiner);
		return true;
	} else {
		return false;
	}
}
 

/**
 * Simple Job: Map -> Reduce -> Map -> Reduce. All functions preserve all fields (hence all properties).
 * 
 * Increases parallelism between 1st reduce and 2nd map, so the hash partitioning from 1st reduce is not reusable.
 * Expected to re-establish partitioning between reduce and map, via hash, because random is a full network
 * transit as well.
 */
@Test
public void checkPropertyHandlingWithIncreasingGlobalParallelism1() {
	final int p = DEFAULT_PARALLELISM;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);
	DataSet<Long> set1 = env.generateSequence(0,1).setParallelism(p);

	set1.map(new IdentityMapper<Long>())
				.withForwardedFields("*").setParallelism(p).name("Map1")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
				.withForwardedFields("*").setParallelism(p).name("Reduce1")
			.map(new IdentityMapper<Long>())
				.withForwardedFields("*").setParallelism(p * 2).name("Map2")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
				.withForwardedFields("*").setParallelism(p * 2).name("Reduce2")
			.output(new DiscardingOutputFormat<Long>()).setParallelism(p * 2).name("Sink");

	Plan plan = env.createProgramPlan();
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);
	
	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();
	
	ShipStrategyType mapIn = map2Node.getInput().getShipStrategy();
	ShipStrategyType redIn = red2Node.getInput().getShipStrategy();
	
	Assert.assertEquals("Invalid ship strategy for an operator.", ShipStrategyType.PARTITION_HASH, mapIn);
	Assert.assertEquals("Invalid ship strategy for an operator.", ShipStrategyType.FORWARD, redIn);
}
 

@Test
public void testDistinctDestroysPartitioningOfNonDistinctFields() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		env.setParallelism(4);
		
		@SuppressWarnings("unchecked")
		DataSet<Tuple2<Long, Long>> data = env.fromElements(new Tuple2<Long, Long>(0L, 0L), new Tuple2<Long, Long>(1L, 1L))
				.map(new IdentityMapper<Tuple2<Long,Long>>()).setParallelism(4);
		
		data.distinct(1)
			.groupBy(0)
			.sum(1)
			.output(new DiscardingOutputFormat<Tuple2<Long, Long>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode combiner = (SingleInputPlanNode) reducer.getInput().getSource();
		SingleInputPlanNode distinctReducer = (SingleInputPlanNode) combiner.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, sink.getInput().getShipStrategy());
		
		// reducer must repartition, because it works on a different field
		assertEquals(ShipStrategyType.PARTITION_HASH, reducer.getInput().getShipStrategy());

		assertEquals(ShipStrategyType.FORWARD, combiner.getInput().getShipStrategy());
		
		// distinct reducer is partitioned
		assertEquals(ShipStrategyType.PARTITION_HASH, distinctReducer.getInput().getShipStrategy());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testPipelineBreakerBroadcastedAllReduce() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		env.getConfig().setExecutionMode(ExecutionMode.PIPELINED);
		env.setParallelism(64);
		
		DataSet<Long> sourceWithMapper = env.generateSequence(1, 10).map(new IdentityMapper<Long>());
		
		DataSet<Long> bcInput1 = sourceWithMapper
									.map(new IdentityMapper<Long>())
									.reduce(new SelectOneReducer<Long>());
		DataSet<Long> bcInput2 = env.generateSequence(1, 10);
		
		DataSet<Long> result = sourceWithMapper
				.map(new IdentityMapper<Long>())
						.withBroadcastSet(bcInput1, "bc1")
						.withBroadcastSet(bcInput2, "bc2");
		
		result.output(new DiscardingOutputFormat<Long>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		SingleInputPlanNode mapper = (SingleInputPlanNode) sink.getInput().getSource();

		assertEquals(TempMode.NONE, mapper.getInput().getTempMode());
		assertEquals(DataExchangeMode.BATCH, mapper.getInput().getDataExchangeMode());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

/**
 * Simple Job: Map -> Reduce -> Map -> Reduce. All functions preserve all fields (hence all properties).
 * 
 * Increases parallelism between 1st reduce and 2nd map, such that more tasks are on one instance.
 * Expected to re-establish partitioning between map and reduce via a local hash.
 */
@Test
public void checkPropertyHandlingWithIncreasingLocalParallelism() {
	final int p = DEFAULT_PARALLELISM * 2;

	// construct the plan
	ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
	env.setParallelism(p);
	DataSet<Long> set1 = env.generateSequence(0,1).setParallelism(p);

	set1.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p).name("Map1")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p).name("Reduce1")
			.map(new IdentityMapper<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Map2")
			.groupBy("*").reduceGroup(new IdentityGroupReducer<Long>())
			.withForwardedFields("*").setParallelism(p * 2).name("Reduce2")
			.output(new DiscardingOutputFormat<Long>()).setParallelism(p * 2).name("Sink");

	Plan plan = env.createProgramPlan();
	// submit the plan to the compiler
	OptimizedPlan oPlan = compileNoStats(plan);
	
	// check the optimized Plan
	// when reducer 1 distributes its data across the instances of map2, it needs to employ a local hash method,
	// because map2 has twice as many instances and key/value pairs with the same key need to be processed by the same
	// mapper respectively reducer
	SinkPlanNode sinkNode = oPlan.getDataSinks().iterator().next();
	SingleInputPlanNode red2Node = (SingleInputPlanNode) sinkNode.getPredecessor();
	SingleInputPlanNode map2Node = (SingleInputPlanNode) red2Node.getPredecessor();
	
	ShipStrategyType mapIn = map2Node.getInput().getShipStrategy();
	ShipStrategyType reduceIn = red2Node.getInput().getShipStrategy();
	
	Assert.assertTrue("Invalid ship strategy for an operator.", 
			(ShipStrategyType.PARTITION_RANDOM ==  mapIn && ShipStrategyType.PARTITION_HASH == reduceIn) || 
			(ShipStrategyType.PARTITION_HASH == mapIn && ShipStrategyType.FORWARD == reduceIn));
}
 

@Test
public void testPartitioningReused() {
	try {
		ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();
		
		@SuppressWarnings("unchecked")
		DataSet<Tuple3<Long, Long, Long>> input = env.fromElements(new Tuple3<Long, Long, Long>(0L, 0L, 0L));
		
		input
			.groupBy(0).sum(1)
			.groupBy(0, 1).sum(2)
			.output(new DiscardingOutputFormat<Tuple3<Long, Long, Long>>());
		
		Plan p = env.createProgramPlan();
		OptimizedPlan op = compileNoStats(p);
		
		SinkPlanNode sink = op.getDataSinks().iterator().next();
		
		SingleInputPlanNode agg2Reducer = (SingleInputPlanNode) sink.getInput().getSource();
		SingleInputPlanNode agg1Reducer = (SingleInputPlanNode) agg2Reducer.getInput().getSource();
		
		assertEquals(ShipStrategyType.FORWARD, agg2Reducer.getInput().getShipStrategy());
		
		assertEquals(ShipStrategyType.PARTITION_HASH, agg1Reducer.getInput().getShipStrategy());
		assertEquals(new FieldList(0), agg1Reducer.getInput().getShipStrategyKeys());
	}
	catch (Exception e) {
		e.printStackTrace();
		fail(e.getMessage());
	}
}
 

@Test
public void testRecordApiWithDirectSoltionSetUpdate() {
	Plan plan = getTestPlan(true, false);
	
	OptimizedPlan oPlan;
	try {
		oPlan = compileNoStats(plan);
	} catch(CompilerException ce) {
		ce.printStackTrace();
		fail("The pact compiler is unable to compile this plan correctly.");
		return; // silence the compiler
	}
	
	OptimizerPlanNodeResolver resolver = getOptimizerPlanNodeResolver(oPlan);
	DualInputPlanNode joinWithInvariantNode = resolver.getNode(JOIN_WITH_INVARIANT_NAME);
	DualInputPlanNode joinWithSolutionSetNode = resolver.getNode(JOIN_WITH_SOLUTION_SET);
	SingleInputPlanNode worksetReducer = resolver.getNode(NEXT_WORKSET_REDUCER_NAME);
	
	// iteration preserves partitioning in reducer, so the first partitioning is out of the loop, 
	// the in-loop partitioning is before the final reducer
	
	// verify joinWithInvariant
	assertEquals(ShipStrategyType.FORWARD, joinWithInvariantNode.getInput1().getShipStrategy()); 
	assertEquals(ShipStrategyType.PARTITION_HASH, joinWithInvariantNode.getInput2().getShipStrategy());
	assertEquals(list0, joinWithInvariantNode.getKeysForInput1());
	assertEquals(list0, joinWithInvariantNode.getKeysForInput2());
	
	// verify joinWithSolutionSet
	assertEquals(ShipStrategyType.FORWARD, joinWithSolutionSetNode.getInput1().getShipStrategy());
	assertEquals(ShipStrategyType.FORWARD, joinWithSolutionSetNode.getInput2().getShipStrategy());
	
	// verify reducer
	assertEquals(ShipStrategyType.FORWARD, worksetReducer.getInput().getShipStrategy());
	assertEquals(list0, worksetReducer.getKeys(0));
	
	
	// verify solution delta
	assertEquals(1, joinWithSolutionSetNode.getOutgoingChannels().size());
	assertEquals(ShipStrategyType.FORWARD, joinWithSolutionSetNode.getOutgoingChannels().get(0).getShipStrategy());
	
	new JobGraphGenerator().compileJobGraph(oPlan);
}
 

@Override
public SingleInputPlanNode instantiate(Channel in, SingleInputNode node) {
	return new SingleInputPlanNode(node, "GroupReduce ("+node.getOperator().getName()+")", in, DriverStrategy.ALL_GROUP_REDUCE);
}
 

@Override
public SingleInputPlanNode instantiate(Channel in, SingleInputNode node) {
	return new SingleInputPlanNode(node, "GroupReduce ("+node.getOperator().getName()+")", in, DriverStrategy.SORTED_GROUP_REDUCE, this.keyList);
}
 

@Override
public SingleInputPlanNode instantiate(Channel in, SingleInputNode node) {
	return new SingleInputPlanNode(node, "MapPartition ("+node.getOperator().getName()+")", in, DriverStrategy.MAP_PARTITION);
}