Examples with FoldingStateDescriptor org.apache.flink.api.common.state.FoldingStateDescriptor used on opensource projects

Example 1 with FoldingStateDescriptor

use of org.apache.flink.api.common.state.FoldingStateDescriptor in project flink by apache.

the class WindowOperatorTest method testCleanupTimerWithEmptyFoldingStateForSessionWindows.

// TODO this test seems invalid, as it uses the unsupported combination of merging windows and folding window state
@Test
public void testCleanupTimerWithEmptyFoldingStateForSessionWindows() throws Exception {
    final int GAP_SIZE = 3;
    final long LATENESS = 10;
    TypeInformation<Tuple2<String, Integer>> inputType = TypeInfoParser.parse("Tuple2<String, Integer>");
    FoldingStateDescriptor<Tuple2<String, Integer>, Tuple2<String, Integer>> windowStateDesc = new FoldingStateDescriptor<>("window-contents", new Tuple2<>((String) null, 0), new FoldFunction<Tuple2<String, Integer>, Tuple2<String, Integer>>() {

        private static final long serialVersionUID = 1L;

        @Override
        public Tuple2<String, Integer> fold(Tuple2<String, Integer> accumulator, Tuple2<String, Integer> value) throws Exception {
            return new Tuple2<>(value.f0, accumulator.f1 + value.f1);
        }
    }, inputType);
    windowStateDesc.initializeSerializerUnlessSet(new ExecutionConfig());
    WindowOperator<String, Tuple2<String, Integer>, Tuple2<String, Integer>, Tuple2<String, Integer>, TimeWindow> operator = new WindowOperator<>(EventTimeSessionWindows.withGap(Time.seconds(GAP_SIZE)), new TimeWindow.Serializer(), new TupleKeySelector(), BasicTypeInfo.STRING_TYPE_INFO.createSerializer(new ExecutionConfig()), windowStateDesc, new InternalSingleValueWindowFunction<>(new PassThroughFunction()), EventTimeTrigger.create(), LATENESS, null);
    OneInputStreamOperatorTestHarness<Tuple2<String, Integer>, Tuple2<String, Integer>> testHarness = new KeyedOneInputStreamOperatorTestHarness<>(operator, new TupleKeySelector(), BasicTypeInfo.STRING_TYPE_INFO);
    testHarness.open();
    ConcurrentLinkedQueue<Object> expected = new ConcurrentLinkedQueue<>();
    testHarness.processElement(new StreamRecord<>(new Tuple2<>("key2", 1), 1000));
    testHarness.processWatermark(new Watermark(4998));
    expected.add(new StreamRecord<>(new Tuple2<>("key2", 1), 3999));
    expected.add(new Watermark(4998));
    testHarness.processWatermark(new Watermark(14600));
    expected.add(new Watermark(14600));
    ConcurrentLinkedQueue<Object> actual = testHarness.getOutput();
    TestHarnessUtil.assertOutputEqualsSorted("Output was not correct.", expected, actual, new Tuple2ResultSortComparator());
    testHarness.close();
}

Example 2 with FoldingStateDescriptor

use of org.apache.flink.api.common.state.FoldingStateDescriptor in project flink by apache.

the class WindowTranslationTest method testFoldWithProcessWindowFunctionEventTime.

@Test
@SuppressWarnings("rawtypes")
public void testFoldWithProcessWindowFunctionEventTime() throws Exception {
    StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setStreamTimeCharacteristic(TimeCharacteristic.IngestionTime);
    DataStream<Tuple2<String, Integer>> source = env.fromElements(Tuple2.of("hello", 1), Tuple2.of("hello", 2));
    DataStream<Tuple2<String, Integer>> window = source.keyBy(new TupleKeySelector()).window(TumblingEventTimeWindows.of(Time.of(1, TimeUnit.SECONDS))).fold(new Tuple3<>("", "", 0), new DummyFolder(), new ProcessWindowFunction<Tuple3<String, String, Integer>, Tuple2<String, Integer>, String, TimeWindow>() {

        private static final long serialVersionUID = 1L;

        @Override
        public void process(String key, Context ctx, Iterable<Tuple3<String, String, Integer>> values, Collector<Tuple2<String, Integer>> out) throws Exception {
            for (Tuple3<String, String, Integer> in : values) {
                out.collect(new Tuple2<>(in.f0, in.f2));
            }
        }
    });
    OneInputTransformation<Tuple2<String, Integer>, Tuple2<String, Integer>> transform = (OneInputTransformation<Tuple2<String, Integer>, Tuple2<String, Integer>>) window.getTransformation();
    OneInputStreamOperator<Tuple2<String, Integer>, Tuple2<String, Integer>> operator = transform.getOperator();
    Assert.assertTrue(operator instanceof WindowOperator);
    WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?> winOperator = (WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?>) operator;
    Assert.assertTrue(winOperator.getTrigger() instanceof EventTimeTrigger);
    Assert.assertTrue(winOperator.getWindowAssigner() instanceof TumblingEventTimeWindows);
    Assert.assertTrue(winOperator.getStateDescriptor() instanceof FoldingStateDescriptor);
    processElementAndEnsureOutput(winOperator, winOperator.getKeySelector(), BasicTypeInfo.STRING_TYPE_INFO, new Tuple2<>("hello", 1));
}

Example 3 with FoldingStateDescriptor

use of org.apache.flink.api.common.state.FoldingStateDescriptor in project flink by apache.

the class WindowTranslationTest method testFoldEventTime.

// ------------------------------------------------------------------------
//  Fold Translation Tests
// ------------------------------------------------------------------------
@Test
@SuppressWarnings("rawtypes")
public void testFoldEventTime() throws Exception {
    StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setStreamTimeCharacteristic(TimeCharacteristic.IngestionTime);
    DataStream<Tuple2<String, Integer>> source = env.fromElements(Tuple2.of("hello", 1), Tuple2.of("hello", 2));
    DataStream<Tuple3<String, String, Integer>> window1 = source.keyBy(0).window(SlidingEventTimeWindows.of(Time.of(1, TimeUnit.SECONDS), Time.of(100, TimeUnit.MILLISECONDS))).fold(new Tuple3<>("", "", 1), new DummyFolder());
    OneInputTransformation<Tuple2<String, Integer>, Tuple3<String, String, Integer>> transform = (OneInputTransformation<Tuple2<String, Integer>, Tuple3<String, String, Integer>>) window1.getTransformation();
    OneInputStreamOperator<Tuple2<String, Integer>, Tuple3<String, String, Integer>> operator = transform.getOperator();
    Assert.assertTrue(operator instanceof WindowOperator);
    WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?> winOperator = (WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?>) operator;
    Assert.assertTrue(winOperator.getTrigger() instanceof EventTimeTrigger);
    Assert.assertTrue(winOperator.getWindowAssigner() instanceof SlidingEventTimeWindows);
    Assert.assertTrue(winOperator.getStateDescriptor() instanceof FoldingStateDescriptor);
    processElementAndEnsureOutput(winOperator, winOperator.getKeySelector(), BasicTypeInfo.STRING_TYPE_INFO, new Tuple2<>("hello", 1));
}

Example 4 with FoldingStateDescriptor

use of org.apache.flink.api.common.state.FoldingStateDescriptor in project flink by apache.

the class WindowTranslationTest method testFoldProcessingTime.

@Test
@SuppressWarnings("rawtypes")
public void testFoldProcessingTime() throws Exception {
    StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
    env.setStreamTimeCharacteristic(TimeCharacteristic.ProcessingTime);
    DataStream<Tuple2<String, Integer>> source = env.fromElements(Tuple2.of("hello", 1), Tuple2.of("hello", 2));
    DataStream<Tuple3<String, String, Integer>> window = source.keyBy(new TupleKeySelector()).window(SlidingProcessingTimeWindows.of(Time.of(1, TimeUnit.SECONDS), Time.of(100, TimeUnit.MILLISECONDS))).fold(new Tuple3<>("", "", 0), new DummyFolder());
    OneInputTransformation<Tuple2<String, Integer>, Tuple3<String, String, Integer>> transform = (OneInputTransformation<Tuple2<String, Integer>, Tuple3<String, String, Integer>>) window.getTransformation();
    OneInputStreamOperator<Tuple2<String, Integer>, Tuple3<String, String, Integer>> operator = transform.getOperator();
    Assert.assertTrue(operator instanceof WindowOperator);
    WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?> winOperator = (WindowOperator<String, Tuple2<String, Integer>, ?, ?, ?>) operator;
    Assert.assertTrue(winOperator.getTrigger() instanceof ProcessingTimeTrigger);
    Assert.assertTrue(winOperator.getWindowAssigner() instanceof SlidingProcessingTimeWindows);
    Assert.assertTrue(winOperator.getStateDescriptor() instanceof FoldingStateDescriptor);
    processElementAndEnsureOutput(winOperator, winOperator.getKeySelector(), BasicTypeInfo.STRING_TYPE_INFO, new Tuple2<>("hello", 1));
}

Example 5 with FoldingStateDescriptor

use of org.apache.flink.api.common.state.FoldingStateDescriptor in project flink by apache.

the class AbstractQueryableStateITCase method testFoldingState.

/**
	 * Tests simple folding state queryable state instance. Each source emits
	 * (subtaskIndex, 0)..(subtaskIndex, numElements) tuples, which are then
	 * queried. The folding state sums these up and maps them to Strings. The
	 * test succeeds after each subtask index is queried with result n*(n+1)/2
	 * (as a String).
	 */
@Test
public void testFoldingState() throws Exception {
    // Config
    final Deadline deadline = TEST_TIMEOUT.fromNow();
    final int numElements = 1024;
    final QueryableStateClient client = new QueryableStateClient(cluster.configuration());
    JobID jobId = null;
    try {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setStateBackend(stateBackend);
        env.setParallelism(NUM_SLOTS);
        // Very important, because cluster is shared between tests and we
        // don't explicitly check that all slots are available before
        // submitting.
        env.setRestartStrategy(RestartStrategies.fixedDelayRestart(Integer.MAX_VALUE, 1000));
        DataStream<Tuple2<Integer, Long>> source = env.addSource(new TestAscendingValueSource(numElements));
        // Folding state
        FoldingStateDescriptor<Tuple2<Integer, Long>, String> foldingState = new FoldingStateDescriptor<>("any", "0", new SumFold(), StringSerializer.INSTANCE);
        QueryableStateStream<Integer, String> queryableState = source.keyBy(new KeySelector<Tuple2<Integer, Long>, Integer>() {

            @Override
            public Integer getKey(Tuple2<Integer, Long> value) throws Exception {
                return value.f0;
            }
        }).asQueryableState("pumba", foldingState);
        // Submit the job graph
        JobGraph jobGraph = env.getStreamGraph().getJobGraph();
        jobId = jobGraph.getJobID();
        cluster.submitJobDetached(jobGraph);
        // Now query
        String expected = Integer.toString(numElements * (numElements + 1) / 2);
        for (int key = 0; key < NUM_SLOTS; key++) {
            final byte[] serializedKey = KvStateRequestSerializer.serializeKeyAndNamespace(key, queryableState.getKeySerializer(), VoidNamespace.INSTANCE, VoidNamespaceSerializer.INSTANCE);
            boolean success = false;
            while (deadline.hasTimeLeft() && !success) {
                Future<byte[]> future = getKvStateWithRetries(client, jobId, queryableState.getQueryableStateName(), key, serializedKey, QUERY_RETRY_DELAY, false);
                byte[] serializedValue = Await.result(future, deadline.timeLeft());
                String value = KvStateRequestSerializer.deserializeValue(serializedValue, queryableState.getValueSerializer());
                if (expected.equals(value)) {
                    success = true;
                } else {
                    // Retry
                    Thread.sleep(50);
                }
            }
            assertTrue("Did not succeed query", success);
        }
    } finally {
        // Free cluster resources
        if (jobId != null) {
            Future<CancellationSuccess> cancellation = cluster.getLeaderGateway(deadline.timeLeft()).ask(new JobManagerMessages.CancelJob(jobId), deadline.timeLeft()).mapTo(ClassTag$.MODULE$.<CancellationSuccess>apply(CancellationSuccess.class));
            Await.ready(cancellation, deadline.timeLeft());
        }
        client.shutDown();
    }
}