Examples with GroupByQueryConfig org.apache.druid.query.groupby.GroupByQueryConfig used on opensource projects

Example 11 with GroupByQueryConfig

use of org.apache.druid.query.groupby.GroupByQueryConfig in project druid by druid-io.

the class GroupByRowProcessor method process.

/**
 * Process the input of sequence "rows" (output by "subquery") based on "query" and returns a {@link ResultSupplier}.
 *
 * In addition to grouping using dimensions and metrics, it will also apply filters (both DimFilter and interval
 * filters).
 *
 * The input sequence is processed synchronously with the call to this method, and result iteration happens lazy upon
 * calls to the {@link ResultSupplier}. Make sure to close it when you're done.
 */
public static ResultSupplier process(final GroupByQuery query, final GroupByQuery subquery, final Sequence<ResultRow> rows, final GroupByQueryConfig config, final GroupByQueryResource resource, final ObjectMapper spillMapper, final String processingTmpDir, final int mergeBufferSize) {
    final Closer closeOnExit = Closer.create();
    final GroupByQueryConfig querySpecificConfig = config.withOverrides(query);
    final File temporaryStorageDirectory = new File(processingTmpDir, StringUtils.format("druid-groupBy-%s_%s", UUID.randomUUID(), query.getId()));
    final LimitedTemporaryStorage temporaryStorage = new LimitedTemporaryStorage(temporaryStorageDirectory, querySpecificConfig.getMaxOnDiskStorage());
    closeOnExit.register(temporaryStorage);
    Pair<Grouper<RowBasedKey>, Accumulator<AggregateResult, ResultRow>> pair = RowBasedGrouperHelper.createGrouperAccumulatorPair(query, subquery, querySpecificConfig, new Supplier<ByteBuffer>() {

        @Override
        public ByteBuffer get() {
            final ResourceHolder<ByteBuffer> mergeBufferHolder = resource.getMergeBuffer();
            closeOnExit.register(mergeBufferHolder);
            return mergeBufferHolder.get();
        }
    }, temporaryStorage, spillMapper, mergeBufferSize);
    final Grouper<RowBasedKey> grouper = pair.lhs;
    final Accumulator<AggregateResult, ResultRow> accumulator = pair.rhs;
    closeOnExit.register(grouper);
    final AggregateResult retVal = rows.accumulate(AggregateResult.ok(), accumulator);
    if (!retVal.isOk()) {
        throw new ResourceLimitExceededException(retVal.getReason());
    }
    return new ResultSupplier() {

        @Override
        public Sequence<ResultRow> results(@Nullable List<DimensionSpec> dimensionsToInclude) {
            return getRowsFromGrouper(query, grouper, dimensionsToInclude);
        }

        @Override
        public void close() throws IOException {
            closeOnExit.close();
        }
    };
}

Example 12 with GroupByQueryConfig

use of org.apache.druid.query.groupby.GroupByQueryConfig in project druid by druid-io.

the class MapVirtualColumnGroupByTest method setup.

@Before
public void setup() throws IOException {
    final IncrementalIndex incrementalIndex = MapVirtualColumnTestBase.generateIndex();
    final GroupByStrategySelector strategySelector = new GroupByStrategySelector(GroupByQueryConfig::new, null, new GroupByStrategyV2(new DruidProcessingConfig() {

        @Override
        public String getFormatString() {
            return null;
        }

        @Override
        public int intermediateComputeSizeBytes() {
            return 10 * 1024 * 1024;
        }

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

        @Override
        public int getNumThreads() {
            return 1;
        }
    }, GroupByQueryConfig::new, new StupidPool<>("map-virtual-column-groupby-test", () -> ByteBuffer.allocate(1024)), new DefaultBlockingPool<>(() -> ByteBuffer.allocate(1024), 1), new DefaultObjectMapper(), QueryRunnerTestHelper.NOOP_QUERYWATCHER));
    final GroupByQueryRunnerFactory factory = new GroupByQueryRunnerFactory(strategySelector, new GroupByQueryQueryToolChest(strategySelector));
    runner = QueryRunnerTestHelper.makeQueryRunner(factory, SegmentId.dummy("index"), new IncrementalIndexSegment(incrementalIndex, SegmentId.dummy("index")), "incremental");
}

Example 13 with GroupByQueryConfig

use of org.apache.druid.query.groupby.GroupByQueryConfig in project druid by druid-io.

the class FixedBucketsHistogramGroupByQueryTest method constructorFeeder.

@Parameterized.Parameters(name = "{0}")
public static Iterable<Object[]> constructorFeeder() {
    final GroupByQueryConfig v1Config = new GroupByQueryConfig() {

        @Override
        public String getDefaultStrategy() {
            return GroupByStrategySelector.STRATEGY_V1;
        }

        @Override
        public String toString() {
            return "v1";
        }
    };
    final GroupByQueryConfig v1SingleThreadedConfig = new GroupByQueryConfig() {

        @Override
        public boolean isSingleThreaded() {
            return true;
        }

        @Override
        public String getDefaultStrategy() {
            return GroupByStrategySelector.STRATEGY_V1;
        }

        @Override
        public String toString() {
            return "v1SingleThreaded";
        }
    };
    final GroupByQueryConfig v2Config = new GroupByQueryConfig() {

        @Override
        public String getDefaultStrategy() {
            return GroupByStrategySelector.STRATEGY_V2;
        }

        @Override
        public String toString() {
            return "v2";
        }
    };
    v1Config.setMaxIntermediateRows(10000);
    v1SingleThreadedConfig.setMaxIntermediateRows(10000);
    final List<Object[]> constructors = new ArrayList<>();
    final List<GroupByQueryConfig> configs = ImmutableList.of(v1Config, v1SingleThreadedConfig, v2Config);
    for (GroupByQueryConfig config : configs) {
        final Pair<GroupByQueryRunnerFactory, Closer> factoryAndCloser = GroupByQueryRunnerTest.makeQueryRunnerFactory(config);
        final GroupByQueryRunnerFactory factory = factoryAndCloser.lhs;
        RESOURCE_CLOSER.register(factoryAndCloser.rhs);
        for (QueryRunner<ResultRow> runner : QueryRunnerTestHelper.makeQueryRunners(factory)) {
            final String testName = StringUtils.format("config=%s, runner=%s", config.toString(), runner.toString());
            constructors.add(new Object[] { testName, factory, runner });
        }
    }
    return constructors;
}

Example 14 with GroupByQueryConfig

use of org.apache.druid.query.groupby.GroupByQueryConfig in project druid by druid-io.

the class CachingClusteredClientBenchmark method setup.

@Setup(Level.Trial)
public void setup() {
    final String schemaName = "basic";
    parallelCombine = parallelism > 0;
    GeneratorSchemaInfo schemaInfo = GeneratorBasicSchemas.SCHEMA_MAP.get(schemaName);
    Map<DataSegment, QueryableIndex> queryableIndexes = Maps.newHashMapWithExpectedSize(numServers);
    for (int i = 0; i < numServers; i++) {
        final DataSegment dataSegment = DataSegment.builder().dataSource(DATA_SOURCE).interval(schemaInfo.getDataInterval()).version("1").shardSpec(new LinearShardSpec(i)).size(0).build();
        final SegmentGenerator segmentGenerator = closer.register(new SegmentGenerator());
        LOG.info("Starting benchmark setup using cacheDir[%s], rows[%,d].", segmentGenerator.getCacheDir(), rowsPerSegment);
        final QueryableIndex index = segmentGenerator.generate(dataSegment, schemaInfo, Granularities.NONE, rowsPerSegment);
        queryableIndexes.put(dataSegment, index);
    }
    final DruidProcessingConfig processingConfig = new DruidProcessingConfig() {

        @Override
        public String getFormatString() {
            return null;
        }

        @Override
        public int intermediateComputeSizeBytes() {
            return PROCESSING_BUFFER_SIZE;
        }

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

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

        @Override
        public boolean useParallelMergePool() {
            return true;
        }
    };
    conglomerate = new DefaultQueryRunnerFactoryConglomerate(ImmutableMap.<Class<? extends Query>, QueryRunnerFactory>builder().put(TimeseriesQuery.class, new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(), new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER)).put(TopNQuery.class, new TopNQueryRunnerFactory(new StupidPool<>("TopNQueryRunnerFactory-bufferPool", () -> ByteBuffer.allocate(PROCESSING_BUFFER_SIZE)), new TopNQueryQueryToolChest(new TopNQueryConfig()), QueryRunnerTestHelper.NOOP_QUERYWATCHER)).put(GroupByQuery.class, makeGroupByQueryRunnerFactory(GroupByQueryRunnerTest.DEFAULT_MAPPER, new GroupByQueryConfig() {

        @Override
        public String getDefaultStrategy() {
            return GroupByStrategySelector.STRATEGY_V2;
        }
    }, processingConfig)).build());
    toolChestWarehouse = new QueryToolChestWarehouse() {

        @Override
        public <T, QueryType extends Query<T>> QueryToolChest<T, QueryType> getToolChest(final QueryType query) {
            return conglomerate.findFactory(query).getToolchest();
        }
    };
    SimpleServerView serverView = new SimpleServerView();
    int serverSuffx = 1;
    for (Entry<DataSegment, QueryableIndex> entry : queryableIndexes.entrySet()) {
        serverView.addServer(createServer(serverSuffx++), entry.getKey(), entry.getValue());
    }
    processingPool = Execs.multiThreaded(processingConfig.getNumThreads(), "caching-clustered-client-benchmark");
    forkJoinPool = new ForkJoinPool((int) Math.ceil(Runtime.getRuntime().availableProcessors() * 0.75), ForkJoinPool.defaultForkJoinWorkerThreadFactory, null, true);
    cachingClusteredClient = new CachingClusteredClient(toolChestWarehouse, serverView, MapCache.create(0), JSON_MAPPER, new ForegroundCachePopulator(JSON_MAPPER, new CachePopulatorStats(), 0), new CacheConfig(), new DruidHttpClientConfig(), processingConfig, forkJoinPool, QueryStackTests.DEFAULT_NOOP_SCHEDULER, new MapJoinableFactory(ImmutableSet.of(), ImmutableMap.of()), new NoopServiceEmitter());
}

Example 15 with GroupByQueryConfig

use of org.apache.druid.query.groupby.GroupByQueryConfig in project druid by druid-io.

the class GroupByTypeInterfaceBenchmark method setup.

@Setup(Level.Trial)
public void setup() throws IOException {
    log.info("SETUP CALLED AT %d", System.currentTimeMillis());
    ComplexMetrics.registerSerde("hyperUnique", new HyperUniquesSerde());
    setupQueries();
    String schemaName = "basic";
    schemaInfo = GeneratorBasicSchemas.SCHEMA_MAP.get(schemaName);
    stringQuery = SCHEMA_QUERY_MAP.get(schemaName).get("string");
    longFloatQuery = SCHEMA_QUERY_MAP.get(schemaName).get("longFloat");
    longQuery = SCHEMA_QUERY_MAP.get(schemaName).get("long");
    floatQuery = SCHEMA_QUERY_MAP.get(schemaName).get("float");
    final DataGenerator dataGenerator = new DataGenerator(schemaInfo.getColumnSchemas(), RNG_SEED + 1, schemaInfo.getDataInterval(), rowsPerSegment);
    tmpDir = FileUtils.createTempDir();
    log.info("Using temp dir: %s", tmpDir.getAbsolutePath());
    // queryableIndexes   -> numSegments worth of on-disk segments
    // anIncrementalIndex -> the last incremental index
    anIncrementalIndex = null;
    queryableIndexes = new ArrayList<>(numSegments);
    for (int i = 0; i < numSegments; i++) {
        log.info("Generating rows for segment %d/%d", i + 1, numSegments);
        final IncrementalIndex index = makeIncIndex();
        for (int j = 0; j < rowsPerSegment; j++) {
            final InputRow row = dataGenerator.nextRow();
            if (j % 20000 == 0) {
                log.info("%,d/%,d rows generated.", i * rowsPerSegment + j, rowsPerSegment * numSegments);
            }
            index.add(row);
        }
        log.info("%,d/%,d rows generated, persisting segment %d/%d.", (i + 1) * rowsPerSegment, rowsPerSegment * numSegments, i + 1, numSegments);
        final File file = INDEX_MERGER_V9.persist(index, new File(tmpDir, String.valueOf(i)), new IndexSpec(), null);
        queryableIndexes.add(INDEX_IO.loadIndex(file));
        if (i == numSegments - 1) {
            anIncrementalIndex = index;
        } else {
            index.close();
        }
    }
    NonBlockingPool<ByteBuffer> bufferPool = new StupidPool<>("GroupByBenchmark-computeBufferPool", new OffheapBufferGenerator("compute", 250_000_000), 0, Integer.MAX_VALUE);
    // limit of 2 is required since we simulate both historical merge and broker merge in the same process
    BlockingPool<ByteBuffer> mergePool = new DefaultBlockingPool<>(new OffheapBufferGenerator("merge", 250_000_000), 2);
    final GroupByQueryConfig config = new GroupByQueryConfig() {

        @Override
        public String getDefaultStrategy() {
            return defaultStrategy;
        }

        @Override
        public int getBufferGrouperInitialBuckets() {
            return initialBuckets;
        }

        @Override
        public long getMaxOnDiskStorage() {
            return 1_000_000_000L;
        }
    };
    config.setSingleThreaded(false);
    config.setMaxIntermediateRows(Integer.MAX_VALUE);
    config.setMaxResults(Integer.MAX_VALUE);
    DruidProcessingConfig druidProcessingConfig = new DruidProcessingConfig() {

        @Override
        public int getNumThreads() {
            // Used by "v2" strategy for concurrencyHint
            return numProcessingThreads;
        }

        @Override
        public String getFormatString() {
            return null;
        }
    };
    final Supplier<GroupByQueryConfig> configSupplier = Suppliers.ofInstance(config);
    final GroupByStrategySelector strategySelector = new GroupByStrategySelector(configSupplier, new GroupByStrategyV1(configSupplier, new GroupByQueryEngine(configSupplier, bufferPool), QueryBenchmarkUtil.NOOP_QUERYWATCHER), new GroupByStrategyV2(druidProcessingConfig, configSupplier, bufferPool, mergePool, new ObjectMapper(new SmileFactory()), QueryBenchmarkUtil.NOOP_QUERYWATCHER));
    factory = new GroupByQueryRunnerFactory(strategySelector, new GroupByQueryQueryToolChest(strategySelector));
}