/** * Tests that the given {@link GroupAlsoByWindowsDoFn} implementation combines elements per * session window correctly according to the provided {@link CombineFn}. */ public static void combinesElementsPerSessionWithEndOfWindowTimestamp( GroupAlsoByWindowsDoFnFactory<String, Long, Long> gabwFactory, CombineFn<Long, ?, Long> combineFn) throws Exception { WindowingStrategy<?, IntervalWindow> windowingStrategy = WindowingStrategy.of(Sessions.withGapDuration(Duration.millis(10))) .withOutputTimeFn(OutputTimeFns.outputAtEndOfWindow()); BoundedWindow secondWindow = window(15, 25); List<WindowedValue<KV<String, Long>>> result = runGABW( gabwFactory, windowingStrategy, "k", WindowedValue.of(1L, new Instant(0), Arrays.asList(window(0, 10)), PaneInfo.NO_FIRING), WindowedValue.of(2L, new Instant(5), Arrays.asList(window(5, 15)), PaneInfo.NO_FIRING), WindowedValue.of(4L, new Instant(15), Arrays.asList(secondWindow), PaneInfo.NO_FIRING)); assertThat(result, hasSize(2)); BoundedWindow firstResultWindow = window(0, 15); TimestampedValue<KV<String, Long>> item0 = getOnlyElementInWindow(result, firstResultWindow); assertThat(item0.getValue().getValue(), equalTo(combineFn.apply(ImmutableList.of(1L, 2L)))); assertThat(item0.getTimestamp(), equalTo(firstResultWindow.maxTimestamp())); TimestampedValue<KV<String, Long>> item1 = getOnlyElementInWindow(result, secondWindow); assertThat(item1.getValue().getValue(), equalTo(combineFn.apply(ImmutableList.of(4L)))); assertThat(item1.getTimestamp(), equalTo(secondWindow.maxTimestamp())); }
/** * Tests that the given GABW implementation correctly groups elements into merged sessions with * output timestamps at the end of the merged window. */ public static void groupsElementsInMergedSessionsWithLatestTimestamp( GroupAlsoByWindowsDoFnFactory<String, String, Iterable<String>> gabwFactory) throws Exception { WindowingStrategy<?, IntervalWindow> windowingStrategy = WindowingStrategy.of(Sessions.withGapDuration(Duration.millis(10))) .withOutputTimeFn(OutputTimeFns.outputAtLatestInputTimestamp()); BoundedWindow unmergedWindow = window(15, 25); List<WindowedValue<KV<String, Iterable<String>>>> result = runGABW( gabwFactory, windowingStrategy, "k", WindowedValue.of( "v1", new Instant(0), Arrays.asList(window(0, 10)), PaneInfo.NO_FIRING), WindowedValue.of( "v2", new Instant(5), Arrays.asList(window(5, 15)), PaneInfo.NO_FIRING), WindowedValue.of( "v3", new Instant(15), Arrays.asList(unmergedWindow), PaneInfo.NO_FIRING)); assertThat(result, hasSize(2)); BoundedWindow mergedWindow = window(0, 15); TimestampedValue<KV<String, Iterable<String>>> item0 = getOnlyElementInWindow(result, mergedWindow); assertThat(item0.getValue().getValue(), containsInAnyOrder("v1", "v2")); assertThat(item0.getTimestamp(), equalTo(new Instant(5))); TimestampedValue<KV<String, Iterable<String>>> item1 = getOnlyElementInWindow(result, unmergedWindow); assertThat(item1.getValue().getValue(), contains("v3")); assertThat(item1.getTimestamp(), equalTo(new Instant(15))); }
/** * Tests that for a simple sequence of elements on the same key, the given GABW implementation * correctly groups them according to fixed windows and also sets the output timestamp according * to the policy {@link OutputTimeFns#outputAtEndOfWindow()}. */ public static void groupsElementsIntoFixedWindowsWithEndOfWindowTimestamp( GroupAlsoByWindowsDoFnFactory<String, String, Iterable<String>> gabwFactory) throws Exception { WindowingStrategy<?, IntervalWindow> windowingStrategy = WindowingStrategy.of(FixedWindows.of(Duration.millis(10))) .withOutputTimeFn(OutputTimeFns.outputAtEndOfWindow()); List<WindowedValue<KV<String, Iterable<String>>>> result = runGABW( gabwFactory, windowingStrategy, "key", WindowedValue.of( "v1", new Instant(1), Arrays.asList(window(0, 10)), PaneInfo.NO_FIRING), WindowedValue.of( "v2", new Instant(2), Arrays.asList(window(0, 10)), PaneInfo.NO_FIRING), WindowedValue.of( "v3", new Instant(13), Arrays.asList(window(10, 20)), PaneInfo.NO_FIRING)); assertThat(result, hasSize(2)); TimestampedValue<KV<String, Iterable<String>>> item0 = getOnlyElementInWindow(result, window(0, 10)); assertThat(item0.getValue().getValue(), containsInAnyOrder("v1", "v2")); assertThat(item0.getTimestamp(), equalTo(window(0, 10).maxTimestamp())); TimestampedValue<KV<String, Iterable<String>>> item1 = getOnlyElementInWindow(result, window(10, 20)); assertThat(item1.getValue().getValue(), contains("v3")); assertThat(item1.getTimestamp(), equalTo(window(10, 20).maxTimestamp())); }
/** * Tests that for a simple sequence of elements on the same key, the given GABW implementation * correctly groups and combines them according to sliding windows. * * <p>In the input here, each element occurs in multiple windows. */ public static void combinesElementsInSlidingWindows( GroupAlsoByWindowsDoFnFactory<String, Long, Long> gabwFactory, CombineFn<Long, ?, Long> combineFn) throws Exception { WindowingStrategy<?, IntervalWindow> windowingStrategy = WindowingStrategy.of(SlidingWindows.of(Duration.millis(20)).every(Duration.millis(10))) .withOutputTimeFn(OutputTimeFns.outputAtEarliestInputTimestamp()); List<WindowedValue<KV<String, Long>>> result = runGABW( gabwFactory, windowingStrategy, "k", WindowedValue.of( 1L, new Instant(5), Arrays.asList(window(-10, 10), window(0, 20)), PaneInfo.NO_FIRING), WindowedValue.of( 2L, new Instant(15), Arrays.asList(window(0, 20), window(10, 30)), PaneInfo.NO_FIRING), WindowedValue.of( 4L, new Instant(18), Arrays.asList(window(0, 20), window(10, 30)), PaneInfo.NO_FIRING)); assertThat(result, hasSize(3)); TimestampedValue<KV<String, Long>> item0 = getOnlyElementInWindow(result, window(-10, 10)); assertThat(item0.getValue().getKey(), equalTo("k")); assertThat(item0.getValue().getValue(), equalTo(combineFn.apply(ImmutableList.of(1L)))); assertThat(item0.getTimestamp(), equalTo(new Instant(5L))); TimestampedValue<KV<String, Long>> item1 = getOnlyElementInWindow(result, window(0, 20)); assertThat(item1.getValue().getKey(), equalTo("k")); assertThat(item1.getValue().getValue(), equalTo(combineFn.apply(ImmutableList.of(1L, 2L, 4L)))); // Timestamp adjusted by WindowFn to exceed the end of the prior sliding window assertThat(item1.getTimestamp(), equalTo(new Instant(10L))); TimestampedValue<KV<String, Long>> item2 = getOnlyElementInWindow(result, window(10, 30)); assertThat(item2.getValue().getKey(), equalTo("k")); assertThat(item2.getValue().getValue(), equalTo(combineFn.apply(ImmutableList.of(2L, 4L)))); // Timestamp adjusted by WindowFn to exceed the end of the prior sliding window assertThat(item2.getTimestamp(), equalTo(new Instant(20L))); }
@Override protected void outputWithTimestampAndWindow( OutputT value, Instant timestamp, Collection<? extends BoundedWindow> windows, PaneInfo pane) { collector.collect(WindowedValue.of(value, timestamp, windows, pane)); }
@Override public void outputWindowedValue( KV<K, Iterable<V>> output, Instant timestamp, Collection<? extends BoundedWindow> windows, PaneInfo pane) { bundle.add(WindowedValue.of(output, timestamp, windows, pane)); }
/** * Tests that for a simple sequence of elements on the same key, the given GABW implementation * correctly groups them into sliding windows. * * <p>In the input here, each element occurs in multiple windows. */ public static void groupsElementsIntoSlidingWindowsWithMinTimestamp( GroupAlsoByWindowsDoFnFactory<String, String, Iterable<String>> gabwFactory) throws Exception { WindowingStrategy<?, IntervalWindow> windowingStrategy = WindowingStrategy.of(SlidingWindows.of(Duration.millis(20)).every(Duration.millis(10))) .withOutputTimeFn(OutputTimeFns.outputAtEarliestInputTimestamp()); List<WindowedValue<KV<String, Iterable<String>>>> result = runGABW( gabwFactory, windowingStrategy, "key", WindowedValue.of( "v1", new Instant(5), Arrays.asList(window(-10, 10), window(0, 20)), PaneInfo.NO_FIRING), WindowedValue.of( "v2", new Instant(15), Arrays.asList(window(0, 20), window(10, 30)), PaneInfo.NO_FIRING)); assertThat(result, hasSize(3)); TimestampedValue<KV<String, Iterable<String>>> item0 = getOnlyElementInWindow(result, window(-10, 10)); assertThat(item0.getValue().getValue(), contains("v1")); assertThat(item0.getTimestamp(), equalTo(new Instant(5))); TimestampedValue<KV<String, Iterable<String>>> item1 = getOnlyElementInWindow(result, window(0, 20)); assertThat(item1.getValue().getValue(), containsInAnyOrder("v1", "v2")); // Timestamp adjusted by WindowFn to exceed the end of the prior sliding window assertThat(item1.getTimestamp(), equalTo(new Instant(10))); TimestampedValue<KV<String, Iterable<String>>> item2 = getOnlyElementInWindow(result, window(10, 30)); assertThat(item2.getValue().getValue(), contains("v2")); // Timestamp adjusted by WindowFn to exceed the end of the prior sliding window assertThat(item2.getTimestamp(), equalTo(new Instant(20))); }