Search in sources :

Example 46 with Node

use of org.apache.beam.runners.dataflow.worker.graph.Nodes.Node in project beam by apache.

the class DeduceNodeLocationsFunctionTest method testGraphWithNonDeducibleNodes.

/**
 * Tests that graphs with deducible and non-deducible nodes are maintained correctly.
 */
@Test
public void testGraphWithNonDeducibleNodes() throws Exception {
    // A --> out1 --\
    // --> Flatten --> D
    // B --> out2 --/-->C
    Node a = createReadNode("A", CUSTOM_SOURCE);
    Node out1 = InstructionOutputNode.create(new InstructionOutput(), "fakeId");
    Node b = createReadNode("B", RUNNER_SOURCE);
    Node out2 = InstructionOutputNode.create(new InstructionOutput(), "fakeId");
    Node c = createParDoNode("C", "RunnerDoFn");
    Node flatten = ParallelInstructionNode.create(new ParallelInstruction().setName("Flatten").setFlatten(new FlattenInstruction()), Nodes.ExecutionLocation.UNKNOWN);
    Node d = createParDoNode("D", DO_FN);
    MutableNetwork<Node, Edge> network = createEmptyNetwork();
    network.addNode(a);
    network.addNode(out1);
    network.addNode(b);
    network.addNode(out2);
    network.addNode(c);
    network.addNode(flatten);
    network.addNode(d);
    network.addEdge(a, out1, DefaultEdge.create());
    network.addEdge(b, out2, DefaultEdge.create());
    network.addEdge(out1, flatten, DefaultEdge.create());
    network.addEdge(out2, flatten, DefaultEdge.create());
    network.addEdge(out2, c, DefaultEdge.create());
    network.addEdge(flatten, d, DefaultEdge.create());
    Network<Node, Edge> inputNetwork = ImmutableNetwork.copyOf(network);
    network = new DeduceNodeLocationsFunction().apply(network);
    assertThatNetworksAreIdentical(inputNetwork, network);
    assertAllNodesDeducedExceptFlattens(network);
}
Also used : ParallelInstruction(com.google.api.services.dataflow.model.ParallelInstruction) InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) Node(org.apache.beam.runners.dataflow.worker.graph.Nodes.Node) InstructionOutput(com.google.api.services.dataflow.model.InstructionOutput) FlattenInstruction(com.google.api.services.dataflow.model.FlattenInstruction) Edge(org.apache.beam.runners.dataflow.worker.graph.Edges.Edge) DefaultEdge(org.apache.beam.runners.dataflow.worker.graph.Edges.DefaultEdge) Test(org.junit.Test)

Example 47 with Node

use of org.apache.beam.runners.dataflow.worker.graph.Nodes.Node in project beam by apache.

the class DeduceNodeLocationsFunctionTest method allPathsWithWrappedNodes.

private List<List<Equivalence.Wrapper<Node>>> allPathsWithWrappedNodes(Network<Node, Edge> network) {
    List<List<Node>> paths = Networks.allPathsFromRootsToLeaves(network);
    List<List<Equivalence.Wrapper<Node>>> wrappedPaths = new ArrayList<>();
    for (List<Node> path : paths) {
        List<Equivalence.Wrapper<Node>> wrappedPath = new ArrayList<>();
        for (Node node : path) {
            wrappedPath.add(NODE_EQUIVALENCE.wrap(node));
        }
        wrappedPaths.add(wrappedPath);
    }
    return wrappedPaths;
}
Also used : Equivalence(org.apache.beam.vendor.guava.v26_0_jre.com.google.common.base.Equivalence) InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) Node(org.apache.beam.runners.dataflow.worker.graph.Nodes.Node) ArrayList(java.util.ArrayList) ArrayList(java.util.ArrayList) List(java.util.List) ImmutableList(org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.ImmutableList)

Example 48 with Node

use of org.apache.beam.runners.dataflow.worker.graph.Nodes.Node in project beam by apache.

the class DeduceNodeLocationsFunctionTest method testSingleNodeWithRunnerRead.

@Test
public void testSingleNodeWithRunnerRead() throws Exception {
    Node unknown = createReadNode("Unknown", RUNNER_SOURCE);
    MutableNetwork<Node, Edge> network = createEmptyNetwork();
    network.addNode(unknown);
    Network<Node, Edge> inputNetwork = ImmutableNetwork.copyOf(network);
    network = new DeduceNodeLocationsFunction().apply(network);
    assertThatNetworksAreIdentical(inputNetwork, network);
    for (Node node : ImmutableList.copyOf(network.nodes())) {
        assertNodesIdenticalExceptForExecutionLocation(unknown, node);
        assertThatLocationIsProperlyDeduced(node, ExecutionLocation.RUNNER_HARNESS);
    }
}
Also used : InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) Node(org.apache.beam.runners.dataflow.worker.graph.Nodes.Node) Edge(org.apache.beam.runners.dataflow.worker.graph.Edges.Edge) DefaultEdge(org.apache.beam.runners.dataflow.worker.graph.Edges.DefaultEdge) Test(org.junit.Test)

Example 49 with Node

use of org.apache.beam.runners.dataflow.worker.graph.Nodes.Node in project beam by apache.

the class InsertFetchAndFilterStreamingSideInputNodesTest method testSdkParDoWithoutSideInput.

@Test
public void testSdkParDoWithoutSideInput() throws Exception {
    Pipeline p = Pipeline.create();
    PCollection<String> pc = p.apply(Create.of("a", "b", "c"));
    pc.apply(ParDo.of(new TestDoFn()));
    RunnerApi.Pipeline pipeline = PipelineTranslation.toProto(p);
    Node predecessor = createParDoNode("predecessor");
    Node mainInput = InstructionOutputNode.create(new InstructionOutput(), "fakeId");
    Node sideInputParDo = createParDoNode("noSideInput");
    MutableNetwork<Node, Edge> network = createEmptyNetwork();
    network.addNode(predecessor);
    network.addNode(mainInput);
    network.addNode(sideInputParDo);
    network.addEdge(predecessor, mainInput, DefaultEdge.create());
    network.addEdge(mainInput, sideInputParDo, DefaultEdge.create());
    Network<Node, Edge> inputNetwork = ImmutableNetwork.copyOf(network);
    network = InsertFetchAndFilterStreamingSideInputNodes.with(pipeline).forNetwork(network);
    assertThatNetworksAreIdentical(inputNetwork, network);
}
Also used : RunnerApi(org.apache.beam.model.pipeline.v1.RunnerApi) FetchAndFilterStreamingSideInputsNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.FetchAndFilterStreamingSideInputsNode) InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) Node(org.apache.beam.runners.dataflow.worker.graph.Nodes.Node) InstructionOutput(com.google.api.services.dataflow.model.InstructionOutput) Edge(org.apache.beam.runners.dataflow.worker.graph.Edges.Edge) DefaultEdge(org.apache.beam.runners.dataflow.worker.graph.Edges.DefaultEdge) Pipeline(org.apache.beam.sdk.Pipeline) Test(org.junit.Test)

Example 50 with Node

use of org.apache.beam.runners.dataflow.worker.graph.Nodes.Node in project beam by apache.

the class MapTaskToNetworkFunctionTest method testParDo.

@Test
public void testParDo() {
    InstructionOutput readOutput = createInstructionOutput("Read.out");
    ParallelInstruction read = createParallelInstruction("Read", readOutput);
    read.setRead(new ReadInstruction());
    MultiOutputInfo parDoMultiOutput = createMultiOutputInfo("output");
    ParDoInstruction parDoInstruction = new ParDoInstruction();
    // Read.out
    parDoInstruction.setInput(createInstructionInput(0, 0));
    parDoInstruction.setMultiOutputInfos(ImmutableList.of(parDoMultiOutput));
    InstructionOutput parDoOutput = createInstructionOutput("ParDo.out");
    ParallelInstruction parDo = createParallelInstruction("ParDo", parDoOutput);
    parDo.setParDo(parDoInstruction);
    MapTask mapTask = new MapTask();
    mapTask.setInstructions(ImmutableList.of(read, parDo));
    mapTask.setFactory(Transport.getJsonFactory());
    Network<Node, Edge> network = new MapTaskToNetworkFunction(IdGenerators.decrementingLongs()).apply(mapTask);
    assertNetworkProperties(network);
    assertEquals(4, network.nodes().size());
    assertEquals(3, network.edges().size());
    ParallelInstructionNode readNode = get(network, read);
    InstructionOutputNode readOutputNode = getOnlySuccessor(network, readNode);
    assertEquals(readOutput, readOutputNode.getInstructionOutput());
    ParallelInstructionNode parDoNode = getOnlySuccessor(network, readOutputNode);
    InstructionOutputNode parDoOutputNode = getOnlySuccessor(network, parDoNode);
    assertEquals(parDoOutput, parDoOutputNode.getInstructionOutput());
    assertEquals(parDoMultiOutput, ((MultiOutputInfoEdge) Iterables.getOnlyElement(network.edgesConnecting(parDoNode, parDoOutputNode))).getMultiOutputInfo());
}
Also used : ParallelInstruction(com.google.api.services.dataflow.model.ParallelInstruction) ParDoInstruction(com.google.api.services.dataflow.model.ParDoInstruction) InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) MultiOutputInfo(com.google.api.services.dataflow.model.MultiOutputInfo) MapTask(com.google.api.services.dataflow.model.MapTask) InstructionOutputNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) Node(org.apache.beam.runners.dataflow.worker.graph.Nodes.Node) InstructionOutput(com.google.api.services.dataflow.model.InstructionOutput) ParallelInstructionNode(org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode) ReadInstruction(com.google.api.services.dataflow.model.ReadInstruction) Edge(org.apache.beam.runners.dataflow.worker.graph.Edges.Edge) DefaultEdge(org.apache.beam.runners.dataflow.worker.graph.Edges.DefaultEdge) MultiOutputInfoEdge(org.apache.beam.runners.dataflow.worker.graph.Edges.MultiOutputInfoEdge) Test(org.junit.Test)

Aggregations

Node (org.apache.beam.runners.dataflow.worker.graph.Nodes.Node)65 ParallelInstructionNode (org.apache.beam.runners.dataflow.worker.graph.Nodes.ParallelInstructionNode)64 InstructionOutputNode (org.apache.beam.runners.dataflow.worker.graph.Nodes.InstructionOutputNode)59 Edge (org.apache.beam.runners.dataflow.worker.graph.Edges.Edge)50 DefaultEdge (org.apache.beam.runners.dataflow.worker.graph.Edges.DefaultEdge)41 Test (org.junit.Test)40 ParallelInstruction (com.google.api.services.dataflow.model.ParallelInstruction)22 MultiOutputInfoEdge (org.apache.beam.runners.dataflow.worker.graph.Edges.MultiOutputInfoEdge)21 InstructionOutput (com.google.api.services.dataflow.model.InstructionOutput)17 List (java.util.List)10 OperationNode (org.apache.beam.runners.dataflow.worker.graph.Nodes.OperationNode)10 MapTask (com.google.api.services.dataflow.model.MapTask)9 ParDoInstruction (com.google.api.services.dataflow.model.ParDoInstruction)9 ReadInstruction (com.google.api.services.dataflow.model.ReadInstruction)9 RemoteGrpcPortNode (org.apache.beam.runners.dataflow.worker.graph.Nodes.RemoteGrpcPortNode)8 FlattenInstruction (com.google.api.services.dataflow.model.FlattenInstruction)7 MultiOutputInfo (com.google.api.services.dataflow.model.MultiOutputInfo)7 ArrayList (java.util.ArrayList)7 HappensBeforeEdge (org.apache.beam.runners.dataflow.worker.graph.Edges.HappensBeforeEdge)7 ParDoOperation (org.apache.beam.runners.dataflow.worker.util.common.worker.ParDoOperation)7