Examples with DataSourceAnalysis org.apache.druid.query.planning.DataSourceAnalysis used on opensource projects

Example 6 with DataSourceAnalysis

use of org.apache.druid.query.planning.DataSourceAnalysis in project druid by druid-io.

the class ClientQuerySegmentWalker method canRunQueryUsingClusterWalker.

/**
 * Checks if a query can be handled wholly by {@link #clusterClient}. Assumes that it is a
 * {@link CachingClusteredClient} or something that behaves similarly.
 */
private <T> boolean canRunQueryUsingClusterWalker(Query<T> query) {
    final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
    final QueryToolChest<T, Query<T>> toolChest = warehouse.getToolChest(query);
    // subqueries on its own).
    return analysis.isConcreteTableBased() && (!analysis.isQuery() || toolChest.canPerformSubquery(((QueryDataSource) analysis.getDataSource()).getQuery()));
}

Example 7 with DataSourceAnalysis

use of org.apache.druid.query.planning.DataSourceAnalysis in project druid by druid-io.

the class ServerManager method getQueryRunnerForSegments.

@Override
public <T> QueryRunner<T> getQueryRunnerForSegments(Query<T> query, Iterable<SegmentDescriptor> specs) {
    final QueryRunnerFactory<T, Query<T>> factory = conglomerate.findFactory(query);
    if (factory == null) {
        final QueryUnsupportedException e = new QueryUnsupportedException(StringUtils.format("Unknown query type, [%s]", query.getClass()));
        log.makeAlert(e, "Error while executing a query[%s]", query.getId()).addData("dataSource", query.getDataSource()).emit();
        throw e;
    }
    final QueryToolChest<T, Query<T>> toolChest = factory.getToolchest();
    final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
    final AtomicLong cpuTimeAccumulator = new AtomicLong(0L);
    final VersionedIntervalTimeline<String, ReferenceCountingSegment> timeline;
    final Optional<VersionedIntervalTimeline<String, ReferenceCountingSegment>> maybeTimeline = segmentManager.getTimeline(analysis);
    // Make sure this query type can handle the subquery, if present.
    if (analysis.isQuery() && !toolChest.canPerformSubquery(((QueryDataSource) analysis.getDataSource()).getQuery())) {
        throw new ISE("Cannot handle subquery: %s", analysis.getDataSource());
    }
    if (maybeTimeline.isPresent()) {
        timeline = maybeTimeline.get();
    } else {
        return new ReportTimelineMissingSegmentQueryRunner<>(Lists.newArrayList(specs));
    }
    // segmentMapFn maps each base Segment into a joined Segment if necessary.
    final Function<SegmentReference, SegmentReference> segmentMapFn = joinableFactoryWrapper.createSegmentMapFn(analysis.getJoinBaseTableFilter().map(Filters::toFilter).orElse(null), analysis.getPreJoinableClauses(), cpuTimeAccumulator, analysis.getBaseQuery().orElse(query));
    // We compute the join cache key here itself so it doesn't need to be re-computed for every segment
    final Optional<byte[]> cacheKeyPrefix = analysis.isJoin() ? joinableFactoryWrapper.computeJoinDataSourceCacheKey(analysis) : Optional.of(StringUtils.EMPTY_BYTES);
    final FunctionalIterable<QueryRunner<T>> queryRunners = FunctionalIterable.create(specs).transformCat(descriptor -> Collections.singletonList(buildQueryRunnerForSegment(query, descriptor, factory, toolChest, timeline, segmentMapFn, cpuTimeAccumulator, cacheKeyPrefix)));
    return CPUTimeMetricQueryRunner.safeBuild(new FinalizeResultsQueryRunner<>(toolChest.mergeResults(factory.mergeRunners(queryProcessingPool, queryRunners)), toolChest), toolChest, emitter, cpuTimeAccumulator, true);
}

Example 8 with DataSourceAnalysis

use of org.apache.druid.query.planning.DataSourceAnalysis in project druid by druid-io.

the class ServerManager method getQueryRunnerForIntervals.

@Override
public <T> QueryRunner<T> getQueryRunnerForIntervals(Query<T> query, Iterable<Interval> intervals) {
    final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
    final VersionedIntervalTimeline<String, ReferenceCountingSegment> timeline;
    final Optional<VersionedIntervalTimeline<String, ReferenceCountingSegment>> maybeTimeline = segmentManager.getTimeline(analysis);
    if (maybeTimeline.isPresent()) {
        timeline = maybeTimeline.get();
    } else {
        // we must find.
        return new NoopQueryRunner<>();
    }
    FunctionalIterable<SegmentDescriptor> segmentDescriptors = FunctionalIterable.create(intervals).transformCat(timeline::lookup).transformCat(holder -> {
        if (holder == null) {
            return null;
        }
        return FunctionalIterable.create(holder.getObject()).transform(partitionChunk -> new SegmentDescriptor(holder.getInterval(), holder.getVersion(), partitionChunk.getChunkNumber()));
    });
    return getQueryRunnerForSegments(query, segmentDescriptors);
}

Example 9 with DataSourceAnalysis

use of org.apache.druid.query.planning.DataSourceAnalysis in project druid by druid-io.

the class SingleTaskBackgroundRunner method getQueryRunnerImpl.

private <T> QueryRunner<T> getQueryRunnerImpl(Query<T> query) {
    QueryRunner<T> queryRunner = null;
    if (runningItem != null) {
        final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
        final Task task = runningItem.getTask();
        if (analysis.getBaseTableDataSource().isPresent() && task.getDataSource().equals(analysis.getBaseTableDataSource().get().getName())) {
            final QueryRunner<T> taskQueryRunner = task.getQueryRunner(query);
            if (taskQueryRunner != null) {
                queryRunner = taskQueryRunner;
            }
        }
    }
    return new SetAndVerifyContextQueryRunner<>(serverConfig, queryRunner == null ? new NoopQueryRunner<>() : queryRunner);
}

Example 10 with DataSourceAnalysis

use of org.apache.druid.query.planning.DataSourceAnalysis in project druid by druid-io.

the class Queries method withBaseDataSource.

/**
 * Rewrite "query" to refer to some specific base datasource, instead of the one it currently refers to.
 *
 * Unlike the seemingly-similar {@link Query#withDataSource}, this will walk down the datasource tree and replace
 * only the base datasource (in the sense defined in {@link DataSourceAnalysis}).
 */
public static <T> Query<T> withBaseDataSource(final Query<T> query, final DataSource newBaseDataSource) {
    final Query<T> retVal;
    if (query.getDataSource() instanceof QueryDataSource) {
        final Query<?> subQuery = ((QueryDataSource) query.getDataSource()).getQuery();
        retVal = query.withDataSource(new QueryDataSource(withBaseDataSource(subQuery, newBaseDataSource)));
    } else {
        final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
        DataSource current = newBaseDataSource;
        DimFilter joinBaseFilter = analysis.getJoinBaseTableFilter().orElse(null);
        for (final PreJoinableClause clause : analysis.getPreJoinableClauses()) {
            current = JoinDataSource.create(current, clause.getDataSource(), clause.getPrefix(), clause.getCondition(), clause.getJoinType(), joinBaseFilter);
            joinBaseFilter = null;
        }
        retVal = query.withDataSource(current);
    }
    // Verify postconditions, just in case.
    final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(retVal.getDataSource());
    if (!newBaseDataSource.equals(analysis.getBaseDataSource())) {
        throw new ISE("Unable to replace base dataSource");
    }
    return retVal;
}