Examples with DimensionTopNMetricSpec io.druid.query.topn.DimensionTopNMetricSpec used on opensource projects

Example 1 with DimensionTopNMetricSpec

use of io.druid.query.topn.DimensionTopNMetricSpec in project druid by druid-io.

the class DruidQueryBuilder method toTopNQuery.

/**
   * Return this query as a TopN query, or null if this query is not compatible with TopN.
   *
   * @param dataSource         data source to query
   * @param sourceRowSignature row signature of the dataSource
   * @param context            query context
   * @param maxTopNLimit       maxTopNLimit from a PlannerConfig
   * @param useApproximateTopN from a PlannerConfig
   *
   * @return query or null
   */
public TopNQuery toTopNQuery(final DataSource dataSource, final RowSignature sourceRowSignature, final Map<String, Object> context, final int maxTopNLimit, final boolean useApproximateTopN) {
    // Must have GROUP BY one column, ORDER BY zero or one column, limit less than maxTopNLimit, and no HAVING.
    final boolean topNOk = grouping != null && grouping.getDimensions().size() == 1 && limitSpec != null && (limitSpec.getColumns().size() <= 1 && limitSpec.getLimit() <= maxTopNLimit) && having == null;
    if (!topNOk) {
        return null;
    }
    final DimensionSpec dimensionSpec = Iterables.getOnlyElement(grouping.getDimensions());
    final OrderByColumnSpec limitColumn;
    if (limitSpec.getColumns().isEmpty()) {
        limitColumn = new OrderByColumnSpec(dimensionSpec.getOutputName(), OrderByColumnSpec.Direction.ASCENDING, StringComparators.LEXICOGRAPHIC);
    } else {
        limitColumn = Iterables.getOnlyElement(limitSpec.getColumns());
    }
    final TopNMetricSpec topNMetricSpec;
    if (limitColumn.getDimension().equals(dimensionSpec.getOutputName())) {
        // DimensionTopNMetricSpec is exact; always return it even if allowApproximate is false.
        final DimensionTopNMetricSpec baseMetricSpec = new DimensionTopNMetricSpec(null, limitColumn.getDimensionComparator());
        topNMetricSpec = limitColumn.getDirection() == OrderByColumnSpec.Direction.ASCENDING ? baseMetricSpec : new InvertedTopNMetricSpec(baseMetricSpec);
    } else if (useApproximateTopN) {
        // ORDER BY metric
        final NumericTopNMetricSpec baseMetricSpec = new NumericTopNMetricSpec(limitColumn.getDimension());
        topNMetricSpec = limitColumn.getDirection() == OrderByColumnSpec.Direction.ASCENDING ? new InvertedTopNMetricSpec(baseMetricSpec) : baseMetricSpec;
    } else {
        return null;
    }
    final Filtration filtration = Filtration.create(filter).optimize(sourceRowSignature);
    return new TopNQuery(dataSource, VirtualColumns.EMPTY, Iterables.getOnlyElement(grouping.getDimensions()), topNMetricSpec, limitSpec.getLimit(), filtration.getQuerySegmentSpec(), filtration.getDimFilter(), Granularities.ALL, grouping.getAggregatorFactories(), grouping.getPostAggregators(), context);
}

Example 2 with DimensionTopNMetricSpec

use of io.druid.query.topn.DimensionTopNMetricSpec in project druid by druid-io.

the class TopNBenchmark method setupQueries.

private void setupQueries() {
    // queries for the basic schema
    Map<String, TopNQueryBuilder> basicQueries = new LinkedHashMap<>();
    BenchmarkSchemaInfo basicSchema = BenchmarkSchemas.SCHEMA_MAP.get("basic");
    {
        // basic.A
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        queryAggs.add(new LongMaxAggregatorFactory("maxLongUniform", "maxLongUniform"));
        queryAggs.add(new DoubleSumAggregatorFactory("sumFloatNormal", "sumFloatNormal"));
        queryAggs.add(new DoubleMinAggregatorFactory("minFloatZipf", "minFloatZipf"));
        queryAggs.add(new HyperUniquesAggregatorFactory("hyperUniquesMet", "hyper"));
        TopNQueryBuilder queryBuilderA = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("dimSequential").metric("sumFloatNormal").intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("A", queryBuilderA);
    }
    {
        // basic.numericSort
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        TopNQueryBuilder queryBuilderA = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("dimUniform").metric(new DimensionTopNMetricSpec(null, StringComparators.NUMERIC)).intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("numericSort", queryBuilderA);
    }
    {
        // basic.alphanumericSort
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        TopNQueryBuilder queryBuilderA = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("dimUniform").metric(new DimensionTopNMetricSpec(null, StringComparators.ALPHANUMERIC)).intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("alphanumericSort", queryBuilderA);
    }
    SCHEMA_QUERY_MAP.put("basic", basicQueries);
}

Example 3 with DimensionTopNMetricSpec

use of io.druid.query.topn.DimensionTopNMetricSpec in project druid by druid-io.

the class TopNTypeInterfaceBenchmark method setupQueries.

private void setupQueries() {
    // queries for the basic schema
    Map<String, TopNQueryBuilder> basicQueries = new LinkedHashMap<>();
    BenchmarkSchemaInfo basicSchema = BenchmarkSchemas.SCHEMA_MAP.get("basic");
    {
        // basic.A
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        queryAggs.add(new LongMaxAggregatorFactory("maxLongUniform", "maxLongUniform"));
        queryAggs.add(new DoubleSumAggregatorFactory("sumFloatNormal", "sumFloatNormal"));
        queryAggs.add(new DoubleMinAggregatorFactory("minFloatZipf", "minFloatZipf"));
        queryAggs.add(new HyperUniquesAggregatorFactory("hyperUniquesMet", "hyper"));
        // Use an IdentityExtractionFn to force usage of DimExtractionTopNAlgorithm
        TopNQueryBuilder queryBuilderString = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension(new ExtractionDimensionSpec("dimSequential", "dimSequential", IdentityExtractionFn.getInstance())).metric("sumFloatNormal").intervals(intervalSpec).aggregators(queryAggs);
        // DimExtractionTopNAlgorithm is always used for numeric columns
        TopNQueryBuilder queryBuilderLong = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("metLongUniform").metric("sumFloatNormal").intervals(intervalSpec).aggregators(queryAggs);
        TopNQueryBuilder queryBuilderFloat = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("metFloatNormal").metric("sumFloatNormal").intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("string", queryBuilderString);
        basicQueries.put("long", queryBuilderLong);
        basicQueries.put("float", queryBuilderFloat);
    }
    {
        // basic.numericSort
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        TopNQueryBuilder queryBuilderA = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("dimUniform").metric(new DimensionTopNMetricSpec(null, StringComparators.NUMERIC)).intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("numericSort", queryBuilderA);
    }
    {
        // basic.alphanumericSort
        QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Arrays.asList(basicSchema.getDataInterval()));
        List<AggregatorFactory> queryAggs = new ArrayList<>();
        queryAggs.add(new LongSumAggregatorFactory("sumLongSequential", "sumLongSequential"));
        TopNQueryBuilder queryBuilderA = new TopNQueryBuilder().dataSource("blah").granularity(Granularities.ALL).dimension("dimUniform").metric(new DimensionTopNMetricSpec(null, StringComparators.ALPHANUMERIC)).intervals(intervalSpec).aggregators(queryAggs);
        basicQueries.put("alphanumericSort", queryBuilderA);
    }
    SCHEMA_QUERY_MAP.put("basic", basicQueries);
}