Examples with IncrementalIndexSegment io.druid.segment.IncrementalIndexSegment used on opensource projects

Example 21 with IncrementalIndexSegment

use of io.druid.segment.IncrementalIndexSegment in project druid by druid-io.

the class OnheapIncrementalIndexBenchmark method testConcurrentAddRead.

@Ignore
@Test
@BenchmarkOptions(callgc = true, clock = Clock.REAL_TIME, warmupRounds = 10, benchmarkRounds = 20)
public void testConcurrentAddRead() throws InterruptedException, ExecutionException, NoSuchMethodException, IllegalAccessException, InvocationTargetException, InstantiationException {
    final int taskCount = 30;
    final int concurrentThreads = 3;
    final int elementsPerThread = 1 << 15;
    final OnheapIncrementalIndex incrementalIndex = this.incrementalIndex.getConstructor(Long.TYPE, Granularity.class, AggregatorFactory[].class, Integer.TYPE).newInstance(0, Granularities.NONE, factories, elementsPerThread * taskCount);
    final ArrayList<AggregatorFactory> queryAggregatorFactories = new ArrayList<>(dimensionCount + 1);
    queryAggregatorFactories.add(new CountAggregatorFactory("rows"));
    for (int i = 0; i < dimensionCount; ++i) {
        queryAggregatorFactories.add(new LongSumAggregatorFactory(String.format("sumResult%s", i), String.format("sumResult%s", i)));
        queryAggregatorFactories.add(new DoubleSumAggregatorFactory(String.format("doubleSumResult%s", i), String.format("doubleSumResult%s", i)));
    }
    final ListeningExecutorService indexExecutor = MoreExecutors.listeningDecorator(Executors.newFixedThreadPool(concurrentThreads, new ThreadFactoryBuilder().setDaemon(false).setNameFormat("index-executor-%d").setPriority(Thread.MIN_PRIORITY).build()));
    final ListeningExecutorService queryExecutor = MoreExecutors.listeningDecorator(Executors.newFixedThreadPool(concurrentThreads, new ThreadFactoryBuilder().setDaemon(false).setNameFormat("query-executor-%d").build()));
    final long timestamp = System.currentTimeMillis();
    final Interval queryInterval = new Interval("1900-01-01T00:00:00Z/2900-01-01T00:00:00Z");
    final List<ListenableFuture<?>> indexFutures = new LinkedList<>();
    final List<ListenableFuture<?>> queryFutures = new LinkedList<>();
    final Segment incrementalIndexSegment = new IncrementalIndexSegment(incrementalIndex, null);
    final QueryRunnerFactory factory = new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator()), new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    final AtomicInteger currentlyRunning = new AtomicInteger(0);
    final AtomicBoolean concurrentlyRan = new AtomicBoolean(false);
    final AtomicBoolean someoneRan = new AtomicBoolean(false);
    for (int j = 0; j < taskCount; j++) {
        indexFutures.add(indexExecutor.submit(new Runnable() {

            @Override
            public void run() {
                currentlyRunning.incrementAndGet();
                try {
                    for (int i = 0; i < elementsPerThread; i++) {
                        incrementalIndex.add(getLongRow(timestamp + i, 1, dimensionCount));
                    }
                } catch (IndexSizeExceededException e) {
                    throw Throwables.propagate(e);
                }
                currentlyRunning.decrementAndGet();
                someoneRan.set(true);
            }
        }));
        queryFutures.add(queryExecutor.submit(new Runnable() {

            @Override
            public void run() {
                QueryRunner<Result<TimeseriesResultValue>> runner = new FinalizeResultsQueryRunner<Result<TimeseriesResultValue>>(factory.createRunner(incrementalIndexSegment), factory.getToolchest());
                TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource("xxx").granularity(Granularities.ALL).intervals(ImmutableList.of(queryInterval)).aggregators(queryAggregatorFactories).build();
                Map<String, Object> context = new HashMap<String, Object>();
                for (Result<TimeseriesResultValue> result : Sequences.toList(runner.run(query, context), new LinkedList<Result<TimeseriesResultValue>>())) {
                    if (someoneRan.get()) {
                        Assert.assertTrue(result.getValue().getDoubleMetric("doubleSumResult0") > 0);
                    }
                }
                if (currentlyRunning.get() > 0) {
                    concurrentlyRan.set(true);
                }
            }
        }));
    }
    List<ListenableFuture<?>> allFutures = new ArrayList<>(queryFutures.size() + indexFutures.size());
    allFutures.addAll(queryFutures);
    allFutures.addAll(indexFutures);
    Futures.allAsList(allFutures).get();
    //Assert.assertTrue("Did not hit concurrency, please try again", concurrentlyRan.get());
    queryExecutor.shutdown();
    indexExecutor.shutdown();
    QueryRunner<Result<TimeseriesResultValue>> runner = new FinalizeResultsQueryRunner<Result<TimeseriesResultValue>>(factory.createRunner(incrementalIndexSegment), factory.getToolchest());
    TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource("xxx").granularity(Granularities.ALL).intervals(ImmutableList.of(queryInterval)).aggregators(queryAggregatorFactories).build();
    Map<String, Object> context = new HashMap<String, Object>();
    List<Result<TimeseriesResultValue>> results = Sequences.toList(runner.run(query, context), new LinkedList<Result<TimeseriesResultValue>>());
    final int expectedVal = elementsPerThread * taskCount;
    for (Result<TimeseriesResultValue> result : results) {
        Assert.assertEquals(elementsPerThread, result.getValue().getLongMetric("rows").intValue());
        for (int i = 0; i < dimensionCount; ++i) {
            Assert.assertEquals(String.format("Failed long sum on dimension %d", i), expectedVal, result.getValue().getLongMetric(String.format("sumResult%s", i)).intValue());
            Assert.assertEquals(String.format("Failed double sum on dimension %d", i), expectedVal, result.getValue().getDoubleMetric(String.format("doubleSumResult%s", i)).intValue());
        }
    }
}

Example 22 with IncrementalIndexSegment

use of io.druid.segment.IncrementalIndexSegment in project druid by druid-io.

the class DistinctCountGroupByQueryTest method testGroupByWithDistinctCountAgg.

@Test
public void testGroupByWithDistinctCountAgg() throws Exception {
    final GroupByQueryConfig config = new GroupByQueryConfig();
    config.setMaxIntermediateRows(10000);
    final GroupByQueryRunnerFactory factory = GroupByQueryRunnerTest.makeQueryRunnerFactory(config);
    IncrementalIndex index = new OnheapIncrementalIndex(0, Granularities.SECOND, new AggregatorFactory[] { new CountAggregatorFactory("cnt") }, 1000);
    String visitor_id = "visitor_id";
    String client_type = "client_type";
    long timestamp = System.currentTimeMillis();
    index.add(new MapBasedInputRow(timestamp, Lists.newArrayList(visitor_id, client_type), ImmutableMap.<String, Object>of(visitor_id, "0", client_type, "iphone")));
    index.add(new MapBasedInputRow(timestamp + 1, Lists.newArrayList(visitor_id, client_type), ImmutableMap.<String, Object>of(visitor_id, "1", client_type, "iphone")));
    index.add(new MapBasedInputRow(timestamp + 2, Lists.newArrayList(visitor_id, client_type), ImmutableMap.<String, Object>of(visitor_id, "2", client_type, "android")));
    GroupByQuery query = new GroupByQuery.Builder().setDataSource(QueryRunnerTestHelper.dataSource).setGranularity(QueryRunnerTestHelper.allGran).setDimensions(Arrays.<DimensionSpec>asList(new DefaultDimensionSpec(client_type, client_type))).setInterval(QueryRunnerTestHelper.fullOnInterval).setLimitSpec(new DefaultLimitSpec(Lists.newArrayList(new OrderByColumnSpec(client_type, OrderByColumnSpec.Direction.DESCENDING)), 10)).setAggregatorSpecs(Lists.newArrayList(QueryRunnerTestHelper.rowsCount, new DistinctCountAggregatorFactory("UV", visitor_id, null))).build();
    final Segment incrementalIndexSegment = new IncrementalIndexSegment(index, null);
    Iterable<Row> results = GroupByQueryRunnerTestHelper.runQuery(factory, factory.createRunner(incrementalIndexSegment), query);
    List<Row> expectedResults = Arrays.asList(GroupByQueryRunnerTestHelper.createExpectedRow("1970-01-01T00:00:00.000Z", client_type, "iphone", "UV", 2L, "rows", 2L), GroupByQueryRunnerTestHelper.createExpectedRow("1970-01-01T00:00:00.000Z", client_type, "android", "UV", 1L, "rows", 1L));
    TestHelper.assertExpectedObjects(expectedResults, results, "distinct-count");
}

Example 23 with IncrementalIndexSegment

use of io.druid.segment.IncrementalIndexSegment in project druid by druid-io.

the class GroupByQueryRunnerFactoryTest method createSegment.

private Segment createSegment() throws Exception {
    IncrementalIndex incrementalIndex = new OnheapIncrementalIndex(0, Granularities.NONE, new AggregatorFactory[] { new CountAggregatorFactory("count") }, true, true, true, 5000);
    StringInputRowParser parser = new StringInputRowParser(new CSVParseSpec(new TimestampSpec("timestamp", "iso", null), new DimensionsSpec(DimensionsSpec.getDefaultSchemas(ImmutableList.of("product", "tags")), null, null), "\t", ImmutableList.of("timestamp", "product", "tags")), "UTF-8");
    String[] rows = new String[] { "2011-01-12T00:00:00.000Z,product_1,t1", "2011-01-13T00:00:00.000Z,product_2,t2", "2011-01-14T00:00:00.000Z,product_3,t2" };
    for (String row : rows) {
        incrementalIndex.add(parser.parse(row));
    }
    closerRule.closeLater(incrementalIndex);
    return new IncrementalIndexSegment(incrementalIndex, "test");
}

Example 24 with IncrementalIndexSegment

use of io.druid.segment.IncrementalIndexSegment in project druid by druid-io.

the class TopNQueryQueryToolChestTest method testMinTopNThreshold.

@Test
public void testMinTopNThreshold() throws Exception {
    TopNQueryConfig config = new TopNQueryConfig();
    final TopNQueryQueryToolChest chest = new TopNQueryQueryToolChest(config, QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator());
    QueryRunnerFactory factory = new TopNQueryRunnerFactory(TestQueryRunners.getPool(), chest, QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    QueryRunner<Result<TopNResultValue>> runner = QueryRunnerTestHelper.makeQueryRunner(factory, new IncrementalIndexSegment(TestIndex.getIncrementalTestIndex(), segmentId), null);
    Map<String, Object> context = Maps.newHashMap();
    context.put("minTopNThreshold", 500);
    TopNQueryBuilder builder = new TopNQueryBuilder().dataSource(QueryRunnerTestHelper.dataSource).granularity(QueryRunnerTestHelper.allGran).dimension(QueryRunnerTestHelper.placementishDimension).metric(QueryRunnerTestHelper.indexMetric).intervals(QueryRunnerTestHelper.fullOnInterval).aggregators(QueryRunnerTestHelper.commonAggregators);
    TopNQuery query1 = builder.threshold(10).context(null).build();
    MockQueryRunner mockRunner = new MockQueryRunner(runner);
    new TopNQueryQueryToolChest.ThresholdAdjustingQueryRunner(mockRunner, config).run(query1, ImmutableMap.<String, Object>of());
    Assert.assertEquals(1000, mockRunner.query.getThreshold());
    TopNQuery query2 = builder.threshold(10).context(context).build();
    new TopNQueryQueryToolChest.ThresholdAdjustingQueryRunner(mockRunner, config).run(query2, ImmutableMap.<String, Object>of());
    Assert.assertEquals(500, mockRunner.query.getThreshold());
    TopNQuery query3 = builder.threshold(2000).context(context).build();
    new TopNQueryQueryToolChest.ThresholdAdjustingQueryRunner(mockRunner, config).run(query3, ImmutableMap.<String, Object>of());
    Assert.assertEquals(2000, mockRunner.query.getThreshold());
}

Example 25 with IncrementalIndexSegment

use of io.druid.segment.IncrementalIndexSegment in project druid by druid-io.

the class TopNQueryRunnerBenchmark method setUp.

@BeforeClass
public static void setUp() throws Exception {
    QueryRunnerFactory factory = new TopNQueryRunnerFactory(new StupidPool<ByteBuffer>("TopNQueryRunnerFactory-directBufferPool", new Supplier<ByteBuffer>() {

        @Override
        public ByteBuffer get() {
            // Instead of causing a circular dependency, we simply mimic its behavior
            return ByteBuffer.allocateDirect(2000);
        }
    }), new TopNQueryQueryToolChest(new TopNQueryConfig(), QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator()), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    testCaseMap.put(TestCases.rtIndex, QueryRunnerTestHelper.makeQueryRunner(factory, new IncrementalIndexSegment(TestIndex.getIncrementalTestIndex(), segmentId), null));
    testCaseMap.put(TestCases.mMappedTestIndex, QueryRunnerTestHelper.makeQueryRunner(factory, new QueryableIndexSegment(segmentId, TestIndex.getMMappedTestIndex()), null));
    testCaseMap.put(TestCases.mergedRealtimeIndex, QueryRunnerTestHelper.makeQueryRunner(factory, new QueryableIndexSegment(segmentId, TestIndex.mergedRealtimeIndex()), null));
//Thread.sleep(10000);
}