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

Example 26 with DoubleSumAggregatorFactory

use of io.druid.query.aggregation.DoubleSumAggregatorFactory 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 27 with DoubleSumAggregatorFactory

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

the class InputRowSerdeTest method testThrowParseExceptions.

@Test(expected = ParseException.class)
public void testThrowParseExceptions() {
    InputRow in = new MapBasedInputRow(timestamp, dims, event);
    AggregatorFactory[] aggregatorFactories = new AggregatorFactory[] { new DoubleSumAggregatorFactory("agg_non_existing", "agg_non_existing_in"), new DoubleSumAggregatorFactory("m1out", "m1"), new LongSumAggregatorFactory("m2out", "m2"), new HyperUniquesAggregatorFactory("m3out", "m3"), // Unparseable from String to Long
    new LongSumAggregatorFactory("unparseable", "m3") };
    InputRowSerde.toBytes(in, aggregatorFactories, true);
}

Example 28 with DoubleSumAggregatorFactory

use of io.druid.query.aggregation.DoubleSumAggregatorFactory in project hive by apache.

the class DruidOutputFormat method getHiveRecordWriter.

@Override
public FileSinkOperator.RecordWriter getHiveRecordWriter(JobConf jc, Path finalOutPath, Class<? extends Writable> valueClass, boolean isCompressed, Properties tableProperties, Progressable progress) throws IOException {
    final String segmentGranularity = tableProperties.getProperty(Constants.DRUID_SEGMENT_GRANULARITY) != null ? tableProperties.getProperty(Constants.DRUID_SEGMENT_GRANULARITY) : HiveConf.getVar(jc, HiveConf.ConfVars.HIVE_DRUID_INDEXING_GRANULARITY);
    final String dataSource = tableProperties.getProperty(Constants.DRUID_DATA_SOURCE);
    final String segmentDirectory = tableProperties.getProperty(Constants.DRUID_SEGMENT_DIRECTORY) != null ? tableProperties.getProperty(Constants.DRUID_SEGMENT_DIRECTORY) : HiveConf.getVar(jc, HiveConf.ConfVars.DRUID_SEGMENT_DIRECTORY);
    final HdfsDataSegmentPusherConfig hdfsDataSegmentPusherConfig = new HdfsDataSegmentPusherConfig();
    hdfsDataSegmentPusherConfig.setStorageDirectory(segmentDirectory);
    final DataSegmentPusher hdfsDataSegmentPusher = new HdfsDataSegmentPusher(hdfsDataSegmentPusherConfig, jc, DruidStorageHandlerUtils.JSON_MAPPER);
    final GranularitySpec granularitySpec = new UniformGranularitySpec(Granularity.valueOf(segmentGranularity), QueryGranularity.fromString(tableProperties.getProperty(Constants.DRUID_QUERY_GRANULARITY) == null ? "NONE" : tableProperties.getProperty(Constants.DRUID_QUERY_GRANULARITY)), null);
    final String columnNameProperty = tableProperties.getProperty(serdeConstants.LIST_COLUMNS);
    final String columnTypeProperty = tableProperties.getProperty(serdeConstants.LIST_COLUMN_TYPES);
    if (StringUtils.isEmpty(columnNameProperty) || StringUtils.isEmpty(columnTypeProperty)) {
        throw new IllegalStateException(String.format("List of columns names [%s] or columns type [%s] is/are not present", columnNameProperty, columnTypeProperty));
    }
    ArrayList<String> columnNames = new ArrayList<String>();
    for (String name : columnNameProperty.split(",")) {
        columnNames.add(name);
    }
    if (!columnNames.contains(DruidTable.DEFAULT_TIMESTAMP_COLUMN)) {
        throw new IllegalStateException("Timestamp column (' " + DruidTable.DEFAULT_TIMESTAMP_COLUMN + "') not specified in create table; list of columns is : " + tableProperties.getProperty(serdeConstants.LIST_COLUMNS));
    }
    ArrayList<TypeInfo> columnTypes = TypeInfoUtils.getTypeInfosFromTypeString(columnTypeProperty);
    // Default, all columns that are not metrics or timestamp, are treated as dimensions
    final List<DimensionSchema> dimensions = new ArrayList<>();
    ImmutableList.Builder<AggregatorFactory> aggregatorFactoryBuilder = ImmutableList.builder();
    for (int i = 0; i < columnTypes.size(); i++) {
        PrimitiveTypeInfo f = (PrimitiveTypeInfo) columnTypes.get(i);
        AggregatorFactory af;
        switch(f.getPrimitiveCategory()) {
            case BYTE:
            case SHORT:
            case INT:
            case LONG:
                af = new LongSumAggregatorFactory(columnNames.get(i), columnNames.get(i));
                break;
            case FLOAT:
            case DOUBLE:
            case DECIMAL:
                af = new DoubleSumAggregatorFactory(columnNames.get(i), columnNames.get(i));
                break;
            case TIMESTAMP:
                String tColumnName = columnNames.get(i);
                if (!tColumnName.equals(DruidTable.DEFAULT_TIMESTAMP_COLUMN) && !tColumnName.equals(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME)) {
                    throw new IOException("Dimension " + tColumnName + " does not have STRING type: " + f.getPrimitiveCategory());
                }
                continue;
            default:
                // Dimension
                String dColumnName = columnNames.get(i);
                if (PrimitiveObjectInspectorUtils.getPrimitiveGrouping(f.getPrimitiveCategory()) != PrimitiveGrouping.STRING_GROUP) {
                    throw new IOException("Dimension " + dColumnName + " does not have STRING type: " + f.getPrimitiveCategory());
                }
                dimensions.add(new StringDimensionSchema(dColumnName));
                continue;
        }
        aggregatorFactoryBuilder.add(af);
    }
    List<AggregatorFactory> aggregatorFactories = aggregatorFactoryBuilder.build();
    final InputRowParser inputRowParser = new MapInputRowParser(new TimeAndDimsParseSpec(new TimestampSpec(DruidTable.DEFAULT_TIMESTAMP_COLUMN, "auto", null), new DimensionsSpec(dimensions, Lists.newArrayList(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME), null)));
    Map<String, Object> inputParser = DruidStorageHandlerUtils.JSON_MAPPER.convertValue(inputRowParser, Map.class);
    final DataSchema dataSchema = new DataSchema(Preconditions.checkNotNull(dataSource, "Data source name is null"), inputParser, aggregatorFactories.toArray(new AggregatorFactory[aggregatorFactories.size()]), granularitySpec, DruidStorageHandlerUtils.JSON_MAPPER);
    final String workingPath = jc.get(Constants.DRUID_JOB_WORKING_DIRECTORY);
    final String version = jc.get(Constants.DRUID_SEGMENT_VERSION);
    Integer maxPartitionSize = HiveConf.getIntVar(jc, HiveConf.ConfVars.HIVE_DRUID_MAX_PARTITION_SIZE);
    String basePersistDirectory = HiveConf.getVar(jc, HiveConf.ConfVars.HIVE_DRUID_BASE_PERSIST_DIRECTORY);
    if (Strings.isNullOrEmpty(basePersistDirectory)) {
        basePersistDirectory = System.getProperty("java.io.tmpdir");
    }
    Integer maxRowInMemory = HiveConf.getIntVar(jc, HiveConf.ConfVars.HIVE_DRUID_MAX_ROW_IN_MEMORY);
    RealtimeTuningConfig realtimeTuningConfig = new RealtimeTuningConfig(maxRowInMemory, null, null, new File(basePersistDirectory, dataSource), new CustomVersioningPolicy(version), null, null, null, null, true, 0, 0, true, null);
    LOG.debug(String.format("running with Data schema [%s] ", dataSchema));
    return new DruidRecordWriter(dataSchema, realtimeTuningConfig, hdfsDataSegmentPusher, maxPartitionSize, new Path(workingPath, SEGMENTS_DESCRIPTOR_DIR_NAME), finalOutPath.getFileSystem(jc));
}

Example 29 with DoubleSumAggregatorFactory

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

the class HadoopConverterJobTest method setUp.

@Before
public void setUp() throws Exception {
    final MetadataStorageUpdaterJobSpec metadataStorageUpdaterJobSpec = new MetadataStorageUpdaterJobSpec() {

        @Override
        public String getSegmentTable() {
            return derbyConnectorRule.metadataTablesConfigSupplier().get().getSegmentsTable();
        }

        @Override
        public MetadataStorageConnectorConfig get() {
            return derbyConnectorRule.getMetadataConnectorConfig();
        }
    };
    final File scratchFileDir = temporaryFolder.newFolder();
    storageLocProperty = System.getProperty(STORAGE_PROPERTY_KEY);
    tmpSegmentDir = temporaryFolder.newFolder();
    System.setProperty(STORAGE_PROPERTY_KEY, tmpSegmentDir.getAbsolutePath());
    final URL url = Preconditions.checkNotNull(Query.class.getClassLoader().getResource("druid.sample.tsv"));
    final File tmpInputFile = temporaryFolder.newFile();
    FileUtils.retryCopy(new ByteSource() {

        @Override
        public InputStream openStream() throws IOException {
            return url.openStream();
        }
    }, tmpInputFile, FileUtils.IS_EXCEPTION, 3);
    final HadoopDruidIndexerConfig hadoopDruidIndexerConfig = new HadoopDruidIndexerConfig(new HadoopIngestionSpec(new DataSchema(DATASOURCE, HadoopDruidIndexerConfig.JSON_MAPPER.convertValue(new StringInputRowParser(new DelimitedParseSpec(new TimestampSpec("ts", "iso", null), new DimensionsSpec(DimensionsSpec.getDefaultSchemas(Arrays.asList(TestIndex.DIMENSIONS)), null, null), "\t", "
", Arrays.asList(TestIndex.COLUMNS)), null), Map.class), new AggregatorFactory[] { new DoubleSumAggregatorFactory(TestIndex.METRICS[0], TestIndex.METRICS[0]), new HyperUniquesAggregatorFactory("quality_uniques", "quality") }, new UniformGranularitySpec(Granularities.MONTH, Granularities.DAY, ImmutableList.<Interval>of(interval)), HadoopDruidIndexerConfig.JSON_MAPPER), new HadoopIOConfig(ImmutableMap.<String, Object>of("type", "static", "paths", tmpInputFile.getAbsolutePath()), metadataStorageUpdaterJobSpec, tmpSegmentDir.getAbsolutePath()), new HadoopTuningConfig(scratchFileDir.getAbsolutePath(), null, null, null, null, null, false, false, false, false, null, false, false, null, null, null, false, false)));
    metadataStorageTablesConfigSupplier = derbyConnectorRule.metadataTablesConfigSupplier();
    connector = derbyConnectorRule.getConnector();
    try {
        connector.getDBI().withHandle(new HandleCallback<Void>() {

            @Override
            public Void withHandle(Handle handle) throws Exception {
                handle.execute("DROP TABLE druid_segments");
                return null;
            }
        });
    } catch (CallbackFailedException e) {
    // Who cares
    }
    List<Jobby> jobs = ImmutableList.of(new Jobby() {

        @Override
        public boolean run() {
            connector.createSegmentTable(metadataStorageUpdaterJobSpec.getSegmentTable());
            return true;
        }
    }, new HadoopDruidDetermineConfigurationJob(hadoopDruidIndexerConfig), new HadoopDruidIndexerJob(hadoopDruidIndexerConfig, new SQLMetadataStorageUpdaterJobHandler(connector)));
    JobHelper.runJobs(jobs, hadoopDruidIndexerConfig);
}

Example 30 with DoubleSumAggregatorFactory

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

the class InputRowSerdeTest method testSerde.

@Test
public void testSerde() {
    InputRow in = new MapBasedInputRow(timestamp, dims, event);
    AggregatorFactory[] aggregatorFactories = new AggregatorFactory[] { new DoubleSumAggregatorFactory("agg_non_existing", "agg_non_existing_in"), new DoubleSumAggregatorFactory("m1out", "m1"), new LongSumAggregatorFactory("m2out", "m2"), new HyperUniquesAggregatorFactory("m3out", "m3"), // Unparseable from String to Long
    new LongSumAggregatorFactory("unparseable", "m3") };
    // Ignore Unparseable aggregator
    byte[] data = InputRowSerde.toBytes(in, aggregatorFactories, false);
    InputRow out = InputRowSerde.fromBytes(data, aggregatorFactories);
    Assert.assertEquals(timestamp, out.getTimestampFromEpoch());
    Assert.assertEquals(dims, out.getDimensions());
    Assert.assertEquals(Collections.EMPTY_LIST, out.getDimension("dim_non_existing"));
    Assert.assertEquals(ImmutableList.of("d1v"), out.getDimension("d1"));
    Assert.assertEquals(ImmutableList.of("d2v1", "d2v2"), out.getDimension("d2"));
    Assert.assertEquals(0.0f, out.getFloatMetric("agg_non_existing"), 0.00001);
    Assert.assertEquals(5.0f, out.getFloatMetric("m1out"), 0.00001);
    Assert.assertEquals(100L, out.getLongMetric("m2out"));
    Assert.assertEquals(1, ((HyperLogLogCollector) out.getRaw("m3out")).estimateCardinality(), 0.001);
    Assert.assertEquals(0L, out.getLongMetric("unparseable"));
}