Examples with DataSchema org.apache.druid.segment.indexing.DataSchema used on opensource projects

Example 76 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class RealtimePlumberSchoolTest method setUp.

@Before
public void setUp() throws Exception {
    tmpDir = FileUtils.createTempDir();
    ObjectMapper jsonMapper = new DefaultObjectMapper();
    schema = new DataSchema("test", jsonMapper.convertValue(new StringInputRowParser(new JSONParseSpec(new TimestampSpec("timestamp", "auto", null), DimensionsSpec.EMPTY, null, null, null), null), Map.class), new AggregatorFactory[] { new CountAggregatorFactory("rows") }, new UniformGranularitySpec(Granularities.HOUR, Granularities.NONE, null), null, jsonMapper);
    schema2 = new DataSchema("test", jsonMapper.convertValue(new StringInputRowParser(new JSONParseSpec(new TimestampSpec("timestamp", "auto", null), DimensionsSpec.EMPTY, null, null, null), null), Map.class), new AggregatorFactory[] { new CountAggregatorFactory("rows") }, new UniformGranularitySpec(Granularities.YEAR, Granularities.NONE, null), null, jsonMapper);
    announcer = EasyMock.createMock(DataSegmentAnnouncer.class);
    announcer.announceSegment(EasyMock.anyObject());
    EasyMock.expectLastCall().anyTimes();
    segmentPublisher = EasyMock.createNiceMock(SegmentPublisher.class);
    dataSegmentPusher = EasyMock.createNiceMock(DataSegmentPusher.class);
    handoffNotifierFactory = EasyMock.createNiceMock(SegmentHandoffNotifierFactory.class);
    handoffNotifier = EasyMock.createNiceMock(SegmentHandoffNotifier.class);
    EasyMock.expect(handoffNotifierFactory.createSegmentHandoffNotifier(EasyMock.anyString())).andReturn(handoffNotifier).anyTimes();
    EasyMock.expect(handoffNotifier.registerSegmentHandoffCallback(EasyMock.anyObject(), EasyMock.anyObject(), EasyMock.anyObject())).andReturn(true).anyTimes();
    emitter = EasyMock.createMock(ServiceEmitter.class);
    EasyMock.replay(announcer, segmentPublisher, dataSegmentPusher, handoffNotifierFactory, handoffNotifier, emitter);
    tuningConfig = new RealtimeTuningConfig(null, 1, null, null, null, null, null, new IntervalStartVersioningPolicy(), rejectionPolicy, null, null, null, null, 0, 0, false, null, null, null, null);
    realtimePlumberSchool = new RealtimePlumberSchool(emitter, new DefaultQueryRunnerFactoryConglomerate(new HashMap<>()), dataSegmentPusher, announcer, segmentPublisher, handoffNotifierFactory, DirectQueryProcessingPool.INSTANCE, NoopJoinableFactory.INSTANCE, TestHelper.getTestIndexMergerV9(segmentWriteOutMediumFactory), TestHelper.getTestIndexIO(), MapCache.create(0), FireDepartmentTest.NO_CACHE_CONFIG, new CachePopulatorStats(), TestHelper.makeJsonMapper());
    metrics = new FireDepartmentMetrics();
    plumber = (RealtimePlumber) realtimePlumberSchool.findPlumber(schema, tuningConfig, metrics);
}

Example 77 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class DruidJsonValidatorTest method testTaskValidator.

@Test
public void testTaskValidator() throws Exception {
    final ObjectMapper jsonMapper = new DefaultObjectMapper();
    for (final Module jacksonModule : new FirehoseModule().getJacksonModules()) {
        jsonMapper.registerModule(jacksonModule);
    }
    final RealtimeIndexTask task = new RealtimeIndexTask(null, new TaskResource("rofl", 2), new FireDepartment(new DataSchema("foo", null, new AggregatorFactory[0], new UniformGranularitySpec(Granularities.HOUR, Granularities.NONE, null), null, jsonMapper), new RealtimeIOConfig(new LocalFirehoseFactory(new File("lol"), "rofl", null), (schema, config, metrics) -> null), new RealtimeTuningConfig(null, 1, null, null, new Period("PT10M"), null, null, null, null, 1, NoneShardSpec.instance(), new IndexSpec(), new IndexSpec(), 0, 0, true, null, null, null, null)), null);
    File tmp = temporaryFolder.newFile("test_task.json");
    jsonMapper.writeValue(tmp, task);
    parseCommand("validator", "-f", tmp.getAbsolutePath(), "-t", "task").run();
}

Example 78 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project druid by druid-io.

the class DefaultOfflineAppenderatorFactoryTest method testBuild.

@Test
public void testBuild() throws IOException, SegmentNotWritableException {
    Injector injector = Initialization.makeInjectorWithModules(GuiceInjectors.makeStartupInjector(), ImmutableList.<Module>of(new Module() {

        @Override
        public void configure(Binder binder) {
            binder.bindConstant().annotatedWith(Names.named("serviceName")).to("druid/tool");
            binder.bindConstant().annotatedWith(Names.named("servicePort")).to(9999);
            binder.bindConstant().annotatedWith(Names.named("tlsServicePort")).to(-1);
            binder.bind(DruidProcessingConfig.class).toInstance(new DruidProcessingConfig() {

                @Override
                public String getFormatString() {
                    return "processing-%s";
                }

                @Override
                public int intermediateComputeSizeBytes() {
                    return 100 * 1024 * 1024;
                }

                @Override
                public int getNumThreads() {
                    return 1;
                }

                @Override
                public int columnCacheSizeBytes() {
                    return 25 * 1024 * 1024;
                }
            });
            binder.bind(ColumnConfig.class).to(DruidProcessingConfig.class);
        }
    }));
    ObjectMapper objectMapper = injector.getInstance(ObjectMapper.class);
    AppenderatorFactory defaultOfflineAppenderatorFactory = objectMapper.readerFor(AppenderatorFactory.class).readValue("{\"type\":\"offline\"}");
    final Map<String, Object> parserMap = objectMapper.convertValue(new MapInputRowParser(new JSONParseSpec(new TimestampSpec("ts", "auto", null), DimensionsSpec.EMPTY, null, null, null)), Map.class);
    DataSchema schema = new DataSchema("dataSourceName", parserMap, new AggregatorFactory[] { new CountAggregatorFactory("count"), new LongSumAggregatorFactory("met", "met") }, new UniformGranularitySpec(Granularities.MINUTE, Granularities.NONE, null), null, objectMapper);
    RealtimeTuningConfig tuningConfig = new RealtimeTuningConfig(null, 75000, null, null, null, null, temporaryFolder.newFolder(), null, null, null, null, null, null, 0, 0, null, null, null, null, null);
    Appenderator appenderator = defaultOfflineAppenderatorFactory.build(schema, tuningConfig, new FireDepartmentMetrics());
    try {
        Assert.assertEquals("dataSourceName", appenderator.getDataSource());
        Assert.assertEquals(null, appenderator.startJob());
        SegmentIdWithShardSpec identifier = new SegmentIdWithShardSpec("dataSourceName", Intervals.of("2000/2001"), "A", new LinearShardSpec(0));
        Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
        appenderator.add(identifier, StreamAppenderatorTest.ir("2000", "bar", 1), null);
        Assert.assertEquals(1, ((AppenderatorImpl) appenderator).getRowsInMemory());
        appenderator.add(identifier, StreamAppenderatorTest.ir("2000", "baz", 1), null);
        Assert.assertEquals(2, ((AppenderatorImpl) appenderator).getRowsInMemory());
        appenderator.close();
        Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
    } finally {
        appenderator.close();
    }
}

Example 79 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema in project hive by apache.

the class DruidStorageHandler method updateKafkaIngestion.

private void updateKafkaIngestion(Table table) {
    final String overlordAddress = HiveConf.getVar(getConf(), HiveConf.ConfVars.HIVE_DRUID_OVERLORD_DEFAULT_ADDRESS);
    final String dataSourceName = Preconditions.checkNotNull(DruidStorageHandlerUtils.getTableProperty(table, Constants.DRUID_DATA_SOURCE), "Druid datasource name is null");
    final String kafkaTopic = Preconditions.checkNotNull(DruidStorageHandlerUtils.getTableProperty(table, DruidConstants.KAFKA_TOPIC), "kafka topic is null");
    final String kafkaServers = Preconditions.checkNotNull(DruidStorageHandlerUtils.getTableProperty(table, DruidConstants.KAFKA_BOOTSTRAP_SERVERS), "kafka connect string is null");
    Properties tableProperties = new Properties();
    tableProperties.putAll(table.getParameters());
    final GranularitySpec granularitySpec = DruidStorageHandlerUtils.getGranularitySpec(getConf(), tableProperties);
    List<FieldSchema> columns = table.getSd().getCols();
    List<String> columnNames = new ArrayList<>(columns.size());
    List<TypeInfo> columnTypes = new ArrayList<>(columns.size());
    for (FieldSchema schema : columns) {
        columnNames.add(schema.getName());
        columnTypes.add(TypeInfoUtils.getTypeInfoFromTypeString(schema.getType()));
    }
    Pair<List<DimensionSchema>, AggregatorFactory[]> dimensionsAndAggregates = DruidStorageHandlerUtils.getDimensionsAndAggregates(columnNames, columnTypes);
    if (!columnNames.contains(DruidConstants.DEFAULT_TIMESTAMP_COLUMN)) {
        throw new IllegalStateException("Timestamp column (' " + DruidConstants.DEFAULT_TIMESTAMP_COLUMN + "') not specified in create table; list of columns is : " + columnNames);
    }
    DimensionsSpec dimensionsSpec = new DimensionsSpec(dimensionsAndAggregates.lhs, null, null);
    String timestampFormat = DruidStorageHandlerUtils.getTableProperty(table, DruidConstants.DRUID_TIMESTAMP_FORMAT);
    String timestampColumnName = DruidStorageHandlerUtils.getTableProperty(table, DruidConstants.DRUID_TIMESTAMP_COLUMN);
    if (timestampColumnName == null) {
        timestampColumnName = DruidConstants.DEFAULT_TIMESTAMP_COLUMN;
    }
    final TimestampSpec timestampSpec = new TimestampSpec(timestampColumnName, timestampFormat, null);
    final InputRowParser inputRowParser = DruidKafkaUtils.getInputRowParser(table, timestampSpec, dimensionsSpec);
    final Map<String, Object> inputParser = JSON_MAPPER.convertValue(inputRowParser, new TypeReference<Map<String, Object>>() {
    });
    final DataSchema dataSchema = new DataSchema(dataSourceName, inputParser, dimensionsAndAggregates.rhs, granularitySpec, null, DruidStorageHandlerUtils.JSON_MAPPER);
    IndexSpec indexSpec = DruidStorageHandlerUtils.getIndexSpec(getConf());
    KafkaSupervisorSpec spec = DruidKafkaUtils.createKafkaSupervisorSpec(table, kafkaTopic, kafkaServers, dataSchema, indexSpec);
    // Fetch existing Ingestion Spec from Druid, if any
    KafkaSupervisorSpec existingSpec = fetchKafkaIngestionSpec(table);
    String targetState = DruidStorageHandlerUtils.getTableProperty(table, DruidConstants.DRUID_KAFKA_INGESTION);
    if (targetState == null) {
        // Case when user has not specified any ingestion state in the current command
        // if there is a kafka supervisor running then keep it last known state is START otherwise STOP.
        targetState = existingSpec == null ? "STOP" : "START";
    }
    if ("STOP".equalsIgnoreCase(targetState)) {
        if (existingSpec != null) {
            stopKafkaIngestion(overlordAddress, dataSourceName);
        }
    } else if ("START".equalsIgnoreCase(targetState)) {
        if (existingSpec == null || !existingSpec.equals(spec)) {
            DruidKafkaUtils.updateKafkaIngestionSpec(overlordAddress, spec);
        }
    } else if ("RESET".equalsIgnoreCase(targetState)) {
        // Case when there are changes in multiple table properties.
        if (existingSpec != null && !existingSpec.equals(spec)) {
            DruidKafkaUtils.updateKafkaIngestionSpec(overlordAddress, spec);
        }
        resetKafkaIngestion(overlordAddress, dataSourceName);
    } else {
        throw new IllegalArgumentException(String.format("Invalid value for property [%s], Valid values are [START, STOP, RESET]", DruidConstants.DRUID_KAFKA_INGESTION));
    }
    // We do not want to keep state in two separate places so remove from hive table properties.
    table.getParameters().remove(DruidConstants.DRUID_KAFKA_INGESTION);
}

Example 80 with DataSchema

use of org.apache.druid.segment.indexing.DataSchema 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 int targetNumShardsPerGranularity = Integer.parseUnsignedInt(tableProperties.getProperty(Constants.DRUID_TARGET_SHARDS_PER_GRANULARITY, "0"));
    final int maxPartitionSize = targetNumShardsPerGranularity > 0 ? -1 : HiveConf.getIntVar(jc, HiveConf.ConfVars.HIVE_DRUID_MAX_PARTITION_SIZE);
    // If datasource is in the table properties, it is an INSERT/INSERT OVERWRITE as the datasource
    // name was already persisted. Otherwise, it is a CT/CTAS and we need to get the name from the
    // job properties that are set by configureOutputJobProperties in the DruidStorageHandler
    final String dataSource = tableProperties.getProperty(Constants.DRUID_DATA_SOURCE) == null ? jc.get(Constants.DRUID_DATA_SOURCE) : tableProperties.getProperty(Constants.DRUID_DATA_SOURCE);
    final String segmentDirectory = jc.get(DruidConstants.DRUID_SEGMENT_INTERMEDIATE_DIRECTORY);
    final GranularitySpec granularitySpec = DruidStorageHandlerUtils.getGranularitySpec(jc, tableProperties);
    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 = Lists.newArrayList(columnNameProperty.split(","));
    if (!columnNames.contains(DruidConstants.DEFAULT_TIMESTAMP_COLUMN)) {
        throw new IllegalStateException("Timestamp column (' " + DruidConstants.DEFAULT_TIMESTAMP_COLUMN + "') not specified in create table; list of columns is : " + tableProperties.getProperty(serdeConstants.LIST_COLUMNS));
    }
    ArrayList<TypeInfo> columnTypes = TypeInfoUtils.getTypeInfosFromTypeString(columnTypeProperty);
    Pair<List<DimensionSchema>, AggregatorFactory[]> dimensionsAndAggregates = DruidStorageHandlerUtils.getDimensionsAndAggregates(columnNames, columnTypes);
    final InputRowParser inputRowParser = new MapInputRowParser(new TimeAndDimsParseSpec(new TimestampSpec(DruidConstants.DEFAULT_TIMESTAMP_COLUMN, "auto", null), new DimensionsSpec(dimensionsAndAggregates.lhs, Lists.newArrayList(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME, Constants.DRUID_SHARD_KEY_COL_NAME), null)));
    Map<String, Object> inputParser = DruidStorageHandlerUtils.JSON_MAPPER.convertValue(inputRowParser, new TypeReference<Map<String, Object>>() {
    });
    final DataSchema dataSchema = new DataSchema(Preconditions.checkNotNull(dataSource, "Data source name is null"), inputParser, dimensionsAndAggregates.rhs, granularitySpec, null, DruidStorageHandlerUtils.JSON_MAPPER);
    final String workingPath = jc.get(DruidConstants.DRUID_JOB_WORKING_DIRECTORY);
    final String version = jc.get(DruidConstants.DRUID_SEGMENT_VERSION);
    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);
    IndexSpec indexSpec = DruidStorageHandlerUtils.getIndexSpec(jc);
    RealtimeTuningConfig realtimeTuningConfig = new RealtimeTuningConfig(maxRowInMemory, null, null, null, new File(basePersistDirectory, dataSource), new CustomVersioningPolicy(version), null, null, null, indexSpec, null, true, 0, 0, true, null, 0L, null, null);
    LOG.debug(String.format("running with Data schema [%s] ", dataSchema));
    return new DruidRecordWriter(dataSchema, realtimeTuningConfig, DruidStorageHandlerUtils.createSegmentPusherForDirectory(segmentDirectory, jc), maxPartitionSize, new Path(workingPath, SEGMENTS_DESCRIPTOR_DIR_NAME), finalOutPath.getFileSystem(jc));
}