Search in sources :

Example 61 with CloudObject

use of org.apache.beam.runners.dataflow.util.CloudObject in project beam by apache.

the class UserParDoFnFactoryTest method testCleanupWorks.

@Test
public void testCleanupWorks() throws Exception {
    PipelineOptions options = PipelineOptionsFactory.create();
    CounterSet counters = new CounterSet();
    DoFn<?, ?> initialFn = new TestStatefulDoFn();
    CloudObject cloudObject = getCloudObject(initialFn, WindowingStrategy.of(FixedWindows.of(Duration.millis(10))));
    StateInternals stateInternals = InMemoryStateInternals.forKey("dummy");
    // The overarching step context that only ParDoFn gets
    DataflowStepContext stepContext = mock(DataflowStepContext.class);
    // The user step context that the DoFnRunner gets a handle on
    DataflowStepContext userStepContext = mock(DataflowStepContext.class);
    when(stepContext.namespacedToUser()).thenReturn(userStepContext);
    when(stepContext.stateInternals()).thenReturn(stateInternals);
    when(userStepContext.stateInternals()).thenReturn((StateInternals) stateInternals);
    DataflowExecutionContext<DataflowStepContext> executionContext = mock(DataflowExecutionContext.class);
    TestOperationContext operationContext = TestOperationContext.create(counters);
    when(executionContext.getStepContext(operationContext)).thenReturn(stepContext);
    when(executionContext.getSideInputReader(any(), any(), any())).thenReturn(NullSideInputReader.empty());
    ParDoFn parDoFn = factory.create(options, cloudObject, Collections.emptyList(), MAIN_OUTPUT, ImmutableMap.of(MAIN_OUTPUT, 0), executionContext, operationContext);
    Receiver rcvr = new OutputReceiver();
    parDoFn.startBundle(rcvr);
    IntervalWindow firstWindow = new IntervalWindow(new Instant(0), new Instant(9));
    IntervalWindow secondWindow = new IntervalWindow(new Instant(10), new Instant(19));
    Coder<IntervalWindow> windowCoder = IntervalWindow.getCoder();
    StateNamespace firstWindowNamespace = StateNamespaces.window(windowCoder, firstWindow);
    StateNamespace secondWindowNamespace = StateNamespaces.window(windowCoder, secondWindow);
    StateTag<ValueState<String>> tag = StateTags.tagForSpec(TestStatefulDoFn.STATE_ID, StateSpecs.value(StringUtf8Coder.of()));
    // Set up non-empty state. We don't mock + verify calls to clear() but instead
    // check that state is actually empty. We musn't care how it is accomplished.
    stateInternals.state(firstWindowNamespace, tag).write("first");
    stateInternals.state(secondWindowNamespace, tag).write("second");
    when(userStepContext.getNextFiredTimer(windowCoder)).thenReturn(null);
    when(stepContext.getNextFiredTimer(windowCoder)).thenReturn(TimerData.of(SimpleParDoFn.CLEANUP_TIMER_ID, firstWindowNamespace, firstWindow.maxTimestamp().plus(Duration.millis(1L)), firstWindow.maxTimestamp().plus(Duration.millis(1L)), TimeDomain.EVENT_TIME)).thenReturn(null);
    // This should fire the timer to clean up the first window
    parDoFn.processTimers();
    assertThat(stateInternals.state(firstWindowNamespace, tag).read(), nullValue());
    assertThat(stateInternals.state(secondWindowNamespace, tag).read(), equalTo("second"));
    when(stepContext.getNextFiredTimer((Coder) windowCoder)).thenReturn(TimerData.of(SimpleParDoFn.CLEANUP_TIMER_ID, secondWindowNamespace, secondWindow.maxTimestamp().plus(Duration.millis(1L)), secondWindow.maxTimestamp().plus(Duration.millis(1L)), TimeDomain.EVENT_TIME)).thenReturn(null);
    // And this should clean up the second window
    parDoFn.processTimers();
    assertThat(stateInternals.state(firstWindowNamespace, tag).read(), nullValue());
    assertThat(stateInternals.state(secondWindowNamespace, tag).read(), nullValue());
}
Also used : Coder(org.apache.beam.sdk.coders.Coder) StringUtf8Coder(org.apache.beam.sdk.coders.StringUtf8Coder) Instant(org.joda.time.Instant) Receiver(org.apache.beam.runners.dataflow.worker.util.common.worker.Receiver) OutputReceiver(org.apache.beam.runners.dataflow.worker.util.common.worker.OutputReceiver) OutputReceiver(org.apache.beam.runners.dataflow.worker.util.common.worker.OutputReceiver) ParDoFn(org.apache.beam.runners.dataflow.worker.util.common.worker.ParDoFn) DataflowStepContext(org.apache.beam.runners.dataflow.worker.DataflowExecutionContext.DataflowStepContext) StateNamespace(org.apache.beam.runners.core.StateNamespace) ValueState(org.apache.beam.sdk.state.ValueState) CounterSet(org.apache.beam.runners.dataflow.worker.counters.CounterSet) CloudObject(org.apache.beam.runners.dataflow.util.CloudObject) PipelineOptions(org.apache.beam.sdk.options.PipelineOptions) InMemoryStateInternals(org.apache.beam.runners.core.InMemoryStateInternals) StateInternals(org.apache.beam.runners.core.StateInternals) IntervalWindow(org.apache.beam.sdk.transforms.windowing.IntervalWindow) Test(org.junit.Test)

Example 62 with CloudObject

use of org.apache.beam.runners.dataflow.util.CloudObject in project beam by apache.

the class UserParDoFnFactoryTest method testFactorySimultaneousUse.

@Test
public void testFactorySimultaneousUse() throws Exception {
    PipelineOptions options = PipelineOptionsFactory.create();
    CounterSet counters = new CounterSet();
    TestDoFn initialFn = new TestDoFn(Collections.<TupleTag<String>>emptyList());
    CloudObject cloudObject = getCloudObject(initialFn);
    ParDoFn parDoFn = factory.create(options, cloudObject, null, MAIN_OUTPUT, ImmutableMap.<TupleTag<?>, Integer>of(MAIN_OUTPUT, 0), BatchModeExecutionContext.forTesting(options, "testStage"), TestOperationContext.create(counters));
    // The fn should not be reused while the first ParDoFn is not finished
    ParDoFn secondParDoFn = factory.create(options, cloudObject, null, MAIN_OUTPUT, ImmutableMap.<TupleTag<?>, Integer>of(MAIN_OUTPUT, 0), BatchModeExecutionContext.forTesting(options, "testStage"), TestOperationContext.create(counters));
    Receiver rcvr = new OutputReceiver();
    parDoFn.startBundle(rcvr);
    parDoFn.processElement(WindowedValue.valueInGlobalWindow("foo"));
    // Must be after the first call to process element for reallyStartBundle to have been called
    TestDoFn firstDoFn = (TestDoFn) ((SimpleParDoFn) parDoFn).getDoFnInfo().getDoFn();
    secondParDoFn.startBundle(rcvr);
    secondParDoFn.processElement(WindowedValue.valueInGlobalWindow("spam"));
    // Must be after the first call to process element for reallyStartBundle to have been called
    TestDoFn secondDoFn = (TestDoFn) ((SimpleParDoFn) secondParDoFn).getDoFnInfo().getDoFn();
    parDoFn.finishBundle();
    secondParDoFn.finishBundle();
    assertThat(firstDoFn, not(theInstance(secondDoFn)));
    assertThat(firstDoFn.state, equalTo(TestDoFn.State.FINISHED));
    assertThat(secondDoFn.state, equalTo(TestDoFn.State.FINISHED));
}
Also used : CounterSet(org.apache.beam.runners.dataflow.worker.counters.CounterSet) CloudObject(org.apache.beam.runners.dataflow.util.CloudObject) PipelineOptions(org.apache.beam.sdk.options.PipelineOptions) Receiver(org.apache.beam.runners.dataflow.worker.util.common.worker.Receiver) OutputReceiver(org.apache.beam.runners.dataflow.worker.util.common.worker.OutputReceiver) OutputReceiver(org.apache.beam.runners.dataflow.worker.util.common.worker.OutputReceiver) ParDoFn(org.apache.beam.runners.dataflow.worker.util.common.worker.ParDoFn) Test(org.junit.Test)

Aggregations

CloudObject (org.apache.beam.runners.dataflow.util.CloudObject)62 ParallelInstruction (com.google.api.services.dataflow.model.ParallelInstruction)23 Test (org.junit.Test)21 Source (com.google.api.services.dataflow.model.Source)15 ParDoInstruction (com.google.api.services.dataflow.model.ParDoInstruction)13 InstructionOutput (com.google.api.services.dataflow.model.InstructionOutput)12 ParDoFn (org.apache.beam.runners.dataflow.worker.util.common.worker.ParDoFn)11 OutputReceiver (org.apache.beam.runners.dataflow.worker.util.common.worker.OutputReceiver)10 PipelineOptions (org.apache.beam.sdk.options.PipelineOptions)10 ByteString (org.apache.beam.vendor.grpc.v1p43p2.com.google.protobuf.ByteString)10 ReadInstruction (com.google.api.services.dataflow.model.ReadInstruction)9 HashMap (java.util.HashMap)9 InstructionInput (com.google.api.services.dataflow.model.InstructionInput)8 Map (java.util.Map)8 ArrayList (java.util.ArrayList)7 Structs.getString (org.apache.beam.runners.dataflow.util.Structs.getString)7 SdkComponents (org.apache.beam.runners.core.construction.SdkComponents)6 Structs.addString (org.apache.beam.runners.dataflow.util.Structs.addString)6 ImmutableList (org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.ImmutableList)6 List (java.util.List)5