Examples with CountAggregatorFactory io.druid.query.aggregation.CountAggregatorFactory used on opensource projects

Example 41 with CountAggregatorFactory

use of io.druid.query.aggregation.CountAggregatorFactory in project druid by druid-io.

the class SqlBenchmark method setup.

@Setup(Level.Trial)
public void setup() throws Exception {
    tmpDir = Files.createTempDir();
    log.info("Starting benchmark setup using tmpDir[%s], rows[%,d].", tmpDir, rowsPerSegment);
    if (ComplexMetrics.getSerdeForType("hyperUnique") == null) {
        ComplexMetrics.registerSerde("hyperUnique", new HyperUniquesSerde(HyperLogLogHash.getDefault()));
    }
    final BenchmarkSchemaInfo schemaInfo = BenchmarkSchemas.SCHEMA_MAP.get("basic");
    final BenchmarkDataGenerator dataGenerator = new BenchmarkDataGenerator(schemaInfo.getColumnSchemas(), RNG_SEED + 1, schemaInfo.getDataInterval(), rowsPerSegment);
    final List<InputRow> rows = Lists.newArrayList();
    for (int i = 0; i < rowsPerSegment; i++) {
        final InputRow row = dataGenerator.nextRow();
        if (i % 20000 == 0) {
            log.info("%,d/%,d rows generated.", i, rowsPerSegment);
        }
        rows.add(row);
    }
    log.info("%,d/%,d rows generated.", rows.size(), rowsPerSegment);
    final PlannerConfig plannerConfig = new PlannerConfig();
    final QueryRunnerFactoryConglomerate conglomerate = CalciteTests.queryRunnerFactoryConglomerate();
    final QueryableIndex index = IndexBuilder.create().tmpDir(new File(tmpDir, "1")).indexMerger(TestHelper.getTestIndexMergerV9()).rows(rows).buildMMappedIndex();
    this.walker = new SpecificSegmentsQuerySegmentWalker(conglomerate).add(DataSegment.builder().dataSource("foo").interval(index.getDataInterval()).version("1").shardSpec(new LinearShardSpec(0)).build(), index);
    final Map<String, Table> tableMap = ImmutableMap.<String, Table>of("foo", new DruidTable(new TableDataSource("foo"), RowSignature.builder().add("__time", ValueType.LONG).add("dimSequential", ValueType.STRING).add("dimZipf", ValueType.STRING).add("dimUniform", ValueType.STRING).build()));
    final Schema druidSchema = new AbstractSchema() {

        @Override
        protected Map<String, Table> getTableMap() {
            return tableMap;
        }
    };
    plannerFactory = new PlannerFactory(Calcites.createRootSchema(druidSchema), walker, CalciteTests.createOperatorTable(), plannerConfig);
    groupByQuery = GroupByQuery.builder().setDataSource("foo").setInterval(new Interval(JodaUtils.MIN_INSTANT, JodaUtils.MAX_INSTANT)).setDimensions(Arrays.<DimensionSpec>asList(new DefaultDimensionSpec("dimZipf", "d0"), new DefaultDimensionSpec("dimSequential", "d1"))).setAggregatorSpecs(Arrays.<AggregatorFactory>asList(new CountAggregatorFactory("c"))).setGranularity(Granularities.ALL).build();
    sqlQuery = "SELECT\n" + "  dimZipf AS d0," + "  dimSequential AS d1,\n" + "  COUNT(*) AS c\n" + "FROM druid.foo\n" + "GROUP BY dimZipf, dimSequential";
}

Example 42 with CountAggregatorFactory

use of io.druid.query.aggregation.CountAggregatorFactory in project druid by druid-io.

the class IncrementalIndexTest method testgetDimensions.

@Test
public void testgetDimensions() {
    final IncrementalIndex<Aggregator> incrementalIndex = new OnheapIncrementalIndex(new IncrementalIndexSchema.Builder().withQueryGranularity(Granularities.NONE).withMetrics(new AggregatorFactory[] { new CountAggregatorFactory("count") }).withDimensionsSpec(new DimensionsSpec(DimensionsSpec.getDefaultSchemas(Arrays.asList("dim0", "dim1")), null, null)).build(), true, 1000000);
    closer.closeLater(incrementalIndex);
    Assert.assertEquals(Arrays.asList("dim0", "dim1"), incrementalIndex.getDimensionNames());
}

Example 43 with CountAggregatorFactory

use of io.druid.query.aggregation.CountAggregatorFactory in project druid by druid-io.

the class IncrementalIndexTest method testSingleThreadedIndexingAndQuery.

@Test
public void testSingleThreadedIndexingAndQuery() throws Exception {
    final int dimensionCount = 5;
    final ArrayList<AggregatorFactory> ingestAggregatorFactories = new ArrayList<>();
    ingestAggregatorFactories.add(new CountAggregatorFactory("rows"));
    for (int i = 0; i < dimensionCount; ++i) {
        ingestAggregatorFactories.add(new LongSumAggregatorFactory(String.format("sumResult%s", i), String.format("Dim_%s", i)));
        ingestAggregatorFactories.add(new DoubleSumAggregatorFactory(String.format("doubleSumResult%s", i), String.format("Dim_%s", i)));
    }
    final IncrementalIndex index = closer.closeLater(indexCreator.createIndex(ingestAggregatorFactories.toArray(new AggregatorFactory[ingestAggregatorFactories.size()])));
    final long timestamp = System.currentTimeMillis();
    final int rows = 50;
    //ingesting same data twice to have some merging happening
    for (int i = 0; i < rows; i++) {
        index.add(getLongRow(timestamp + i, i, dimensionCount));
    }
    for (int i = 0; i < rows; i++) {
        index.add(getLongRow(timestamp + i, i, dimensionCount));
    }
    //run a timeseries query on the index and verify results
    final ArrayList<AggregatorFactory> queryAggregatorFactories = new ArrayList<>();
    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)));
    }
    TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource("xxx").granularity(Granularities.ALL).intervals(ImmutableList.of(new Interval("2000/2030"))).aggregators(queryAggregatorFactories).build();
    final Segment incrementalIndexSegment = new IncrementalIndexSegment(index, null);
    final QueryRunnerFactory factory = new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator()), new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    final QueryRunner<Result<TimeseriesResultValue>> runner = new FinalizeResultsQueryRunner<Result<TimeseriesResultValue>>(factory.createRunner(incrementalIndexSegment), factory.getToolchest());
    List<Result<TimeseriesResultValue>> results = Sequences.toList(runner.run(query, new HashMap<String, Object>()), new LinkedList<Result<TimeseriesResultValue>>());
    Result<TimeseriesResultValue> result = Iterables.getOnlyElement(results);
    boolean isRollup = index.isRollup();
    Assert.assertEquals(rows * (isRollup ? 1 : 2), result.getValue().getLongMetric("rows").intValue());
    for (int i = 0; i < dimensionCount; ++i) {
        Assert.assertEquals(String.format("Failed long sum on dimension %d", i), 2 * rows, result.getValue().getLongMetric(String.format("sumResult%s", i)).intValue());
        Assert.assertEquals(String.format("Failed double sum on dimension %d", i), 2 * rows, result.getValue().getDoubleMetric(String.format("doubleSumResult%s", i)).intValue());
    }
}

Example 44 with CountAggregatorFactory

use of io.druid.query.aggregation.CountAggregatorFactory in project druid by druid-io.

the class IncrementalIndexTest method testConcurrentAddRead.

@Test(timeout = 60_000L)
public void testConcurrentAddRead() throws InterruptedException, ExecutionException {
    final int dimensionCount = 5;
    final ArrayList<AggregatorFactory> ingestAggregatorFactories = new ArrayList<>(dimensionCount + 1);
    ingestAggregatorFactories.add(new CountAggregatorFactory("rows"));
    for (int i = 0; i < dimensionCount; ++i) {
        ingestAggregatorFactories.add(new LongSumAggregatorFactory(String.format("sumResult%s", i), String.format("Dim_%s", i)));
        ingestAggregatorFactories.add(new DoubleSumAggregatorFactory(String.format("doubleSumResult%s", i), String.format("Dim_%s", i)));
    }
    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 IncrementalIndex index = closer.closeLater(indexCreator.createIndex(ingestAggregatorFactories.toArray(new AggregatorFactory[dimensionCount])));
    final int concurrentThreads = 2;
    final int elementsPerThread = 10_000;
    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 = Lists.newArrayListWithExpectedSize(concurrentThreads);
    final List<ListenableFuture<?>> queryFutures = Lists.newArrayListWithExpectedSize(concurrentThreads);
    final Segment incrementalIndexSegment = new IncrementalIndexSegment(index, null);
    final QueryRunnerFactory factory = new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(QueryRunnerTestHelper.NoopIntervalChunkingQueryRunnerDecorator()), new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER);
    final AtomicInteger currentlyRunning = new AtomicInteger(0);
    final AtomicInteger concurrentlyRan = new AtomicInteger(0);
    final AtomicInteger someoneRan = new AtomicInteger(0);
    final CountDownLatch startLatch = new CountDownLatch(1);
    final CountDownLatch readyLatch = new CountDownLatch(concurrentThreads * 2);
    final AtomicInteger queriesAccumualted = new AtomicInteger(0);
    for (int j = 0; j < concurrentThreads; j++) {
        indexFutures.add(indexExecutor.submit(new Runnable() {

            @Override
            public void run() {
                readyLatch.countDown();
                try {
                    startLatch.await();
                } catch (InterruptedException e) {
                    Thread.currentThread().interrupt();
                    throw Throwables.propagate(e);
                }
                currentlyRunning.incrementAndGet();
                try {
                    for (int i = 0; i < elementsPerThread; i++) {
                        index.add(getLongRow(timestamp + i, i, dimensionCount));
                        someoneRan.incrementAndGet();
                    }
                } catch (IndexSizeExceededException e) {
                    throw Throwables.propagate(e);
                }
                currentlyRunning.decrementAndGet();
            }
        }));
        final TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource("xxx").granularity(Granularities.ALL).intervals(ImmutableList.of(queryInterval)).aggregators(queryAggregatorFactories).build();
        queryFutures.add(queryExecutor.submit(new Runnable() {

            @Override
            public void run() {
                readyLatch.countDown();
                try {
                    startLatch.await();
                } catch (InterruptedException e) {
                    Thread.currentThread().interrupt();
                    throw Throwables.propagate(e);
                }
                while (concurrentlyRan.get() == 0) {
                    QueryRunner<Result<TimeseriesResultValue>> runner = new FinalizeResultsQueryRunner<Result<TimeseriesResultValue>>(factory.createRunner(incrementalIndexSegment), factory.getToolchest());
                    Map<String, Object> context = new HashMap<String, Object>();
                    Sequence<Result<TimeseriesResultValue>> sequence = runner.run(query, context);
                    for (Double result : sequence.accumulate(new Double[0], new Accumulator<Double[], Result<TimeseriesResultValue>>() {

                        @Override
                        public Double[] accumulate(Double[] accumulated, Result<TimeseriesResultValue> in) {
                            if (currentlyRunning.get() > 0) {
                                concurrentlyRan.incrementAndGet();
                            }
                            queriesAccumualted.incrementAndGet();
                            return Lists.asList(in.getValue().getDoubleMetric("doubleSumResult0"), accumulated).toArray(new Double[accumulated.length + 1]);
                        }
                    })) {
                        final Integer maxValueExpected = someoneRan.get() + concurrentThreads;
                        if (maxValueExpected > 0) {
                            // Eventually consistent, but should be somewhere in that range
                            // Actual result is validated after all writes are guaranteed done.
                            Assert.assertTrue(String.format("%d >= %g >= 0 violated", maxValueExpected, result), result >= 0 && result <= maxValueExpected);
                        }
                    }
                }
            }
        }));
    }
    readyLatch.await();
    startLatch.countDown();
    List<ListenableFuture<?>> allFutures = new ArrayList<>(queryFutures.size() + indexFutures.size());
    allFutures.addAll(queryFutures);
    allFutures.addAll(indexFutures);
    Futures.allAsList(allFutures).get();
    Assert.assertTrue("Queries ran too fast", queriesAccumualted.get() > 0);
    Assert.assertTrue("Did not hit concurrency, please try again", concurrentlyRan.get() > 0);
    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>>());
    boolean isRollup = index.isRollup();
    for (Result<TimeseriesResultValue> result : results) {
        Assert.assertEquals(elementsPerThread * (isRollup ? 1 : concurrentThreads), result.getValue().getLongMetric("rows").intValue());
        for (int i = 0; i < dimensionCount; ++i) {
            Assert.assertEquals(String.format("Failed long sum on dimension %d", i), elementsPerThread * concurrentThreads, result.getValue().getLongMetric(String.format("sumResult%s", i)).intValue());
            Assert.assertEquals(String.format("Failed double sum on dimension %d", i), elementsPerThread * concurrentThreads, result.getValue().getDoubleMetric(String.format("doubleSumResult%s", i)).intValue());
        }
    }
}

Example 45 with CountAggregatorFactory

use of io.druid.query.aggregation.CountAggregatorFactory in project druid by druid-io.

the class IndexMergerTest method testMergeSpecChange.

@Test
public void testMergeSpecChange() throws Exception {
    final long timestamp = System.currentTimeMillis();
    IncrementalIndex toPersist1 = IncrementalIndexTest.createIndex(null);
    IncrementalIndexTest.populateIndex(timestamp, toPersist1);
    final File tempDir1 = temporaryFolder.newFolder();
    final File mergedDir = temporaryFolder.newFolder();
    final IndexableAdapter incrementalAdapter = new IncrementalIndexAdapter(toPersist1.getInterval(), toPersist1, indexSpec.getBitmapSerdeFactory().getBitmapFactory());
    QueryableIndex index1 = closer.closeLater(INDEX_IO.loadIndex(INDEX_MERGER.persist(toPersist1, tempDir1, indexSpec)));
    final IndexableAdapter queryableAdapter = new QueryableIndexIndexableAdapter(index1);
    INDEX_IO.validateTwoSegments(incrementalAdapter, queryableAdapter);
    Assert.assertEquals(2, index1.getColumn(Column.TIME_COLUMN_NAME).getLength());
    Assert.assertEquals(Arrays.asList("dim1", "dim2"), Lists.newArrayList(index1.getAvailableDimensions()));
    Assert.assertEquals(3, index1.getColumnNames().size());
    IndexSpec newSpec = new IndexSpec(indexSpec.getBitmapSerdeFactory(), CompressedObjectStrategy.CompressionStrategy.LZ4.equals(indexSpec.getDimensionCompression()) ? CompressedObjectStrategy.CompressionStrategy.LZF : CompressedObjectStrategy.CompressionStrategy.LZ4, CompressedObjectStrategy.CompressionStrategy.LZ4.equals(indexSpec.getDimensionCompression()) ? CompressedObjectStrategy.CompressionStrategy.LZF : CompressedObjectStrategy.CompressionStrategy.LZ4, CompressionFactory.LongEncodingStrategy.LONGS.equals(indexSpec.getLongEncoding()) ? CompressionFactory.LongEncodingStrategy.AUTO : CompressionFactory.LongEncodingStrategy.LONGS);
    AggregatorFactory[] mergedAggregators = new AggregatorFactory[] { new CountAggregatorFactory("count") };
    QueryableIndex merged = closer.closeLater(INDEX_IO.loadIndex(INDEX_MERGER.mergeQueryableIndex(ImmutableList.of(index1), true, mergedAggregators, mergedDir, newSpec)));
    Assert.assertEquals(2, merged.getColumn(Column.TIME_COLUMN_NAME).getLength());
    Assert.assertEquals(Arrays.asList("dim1", "dim2"), Lists.newArrayList(merged.getAvailableDimensions()));
    Assert.assertEquals(3, merged.getColumnNames().size());
    INDEX_IO.validateTwoSegments(tempDir1, mergedDir);
    assertDimCompression(index1, indexSpec.getDimensionCompression());
    assertDimCompression(merged, newSpec.getDimensionCompression());
}