Examples with VirtualColumn org.apache.druid.segment.VirtualColumn used on opensource projects

Example 1 with VirtualColumn

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

the class CalciteQueryTest method testCountAndAverageByConstantVirtualColumn.

@Test
public void testCountAndAverageByConstantVirtualColumn() throws Exception {
    List<VirtualColumn> virtualColumns;
    List<AggregatorFactory> aggs;
    if (useDefault) {
        aggs = ImmutableList.of(new FilteredAggregatorFactory(new CountAggregatorFactory("a0"), not(selector("v0", null, null))), new LongSumAggregatorFactory("a1:sum", null, "325323", TestExprMacroTable.INSTANCE), new CountAggregatorFactory("a1:count"));
        virtualColumns = ImmutableList.of(expressionVirtualColumn("v0", "'10.1'", ColumnType.STRING));
    } else {
        aggs = ImmutableList.of(new FilteredAggregatorFactory(new CountAggregatorFactory("a0"), not(selector("v0", null, null))), new LongSumAggregatorFactory("a1:sum", "v1"), new FilteredAggregatorFactory(new CountAggregatorFactory("a1:count"), not(selector("v1", null, null))));
        virtualColumns = ImmutableList.of(expressionVirtualColumn("v0", "'10.1'", ColumnType.STRING), expressionVirtualColumn("v1", "325323", ColumnType.LONG));
    }
    testQuery("SELECT dim5, COUNT(dim1), AVG(l1) FROM druid.numfoo WHERE dim1 = '10.1' AND l1 = 325323 GROUP BY dim5", ImmutableList.of(GroupByQuery.builder().setDataSource(CalciteTests.DATASOURCE3).setInterval(querySegmentSpec(Filtration.eternity())).setDimFilter(and(selector("dim1", "10.1", null), selector("l1", "325323", null))).setGranularity(Granularities.ALL).setVirtualColumns(VirtualColumns.create(virtualColumns)).setDimensions(new DefaultDimensionSpec("dim5", "_d0", ColumnType.STRING)).setAggregatorSpecs(aggs).setPostAggregatorSpecs(ImmutableList.of(new ArithmeticPostAggregator("a1", "quotient", ImmutableList.of(new FieldAccessPostAggregator(null, "a1:sum"), new FieldAccessPostAggregator(null, "a1:count"))))).setContext(QUERY_CONTEXT_DEFAULT).build()), ImmutableList.of(new Object[] { "ab", 1L, 325323L }));
}

Example 2 with VirtualColumn

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

the class JoinFilterAnalyzer method splitFilter.

/**
 * @param joinFilterPreAnalysis The pre-analysis computed by {@link #computeJoinFilterPreAnalysis)}
 * @param baseFilter - Filter on base table that was specified in the query itself
 *
 * @return A JoinFilterSplit indicating what parts of the filter should be applied pre-join and post-join
 */
public static JoinFilterSplit splitFilter(JoinFilterPreAnalysis joinFilterPreAnalysis, @Nullable Filter baseFilter) {
    if (joinFilterPreAnalysis.getOriginalFilter() == null || !joinFilterPreAnalysis.isEnableFilterPushDown()) {
        return new JoinFilterSplit(baseFilter, joinFilterPreAnalysis.getOriginalFilter(), ImmutableSet.of());
    }
    // Pushdown filters, rewriting if necessary
    List<Filter> leftFilters = new ArrayList<>();
    List<Filter> rightFilters = new ArrayList<>();
    Map<Expr, VirtualColumn> pushDownVirtualColumnsForLhsExprs = new HashMap<>();
    if (null != baseFilter) {
        leftFilters.add(baseFilter);
    }
    for (Filter baseTableFilter : joinFilterPreAnalysis.getNormalizedBaseTableClauses()) {
        if (!Filters.filterMatchesNull(baseTableFilter)) {
            leftFilters.add(baseTableFilter);
        } else {
            rightFilters.add(baseTableFilter);
        }
    }
    for (Filter orClause : joinFilterPreAnalysis.getNormalizedJoinTableClauses()) {
        JoinFilterAnalysis joinFilterAnalysis = analyzeJoinFilterClause(orClause, joinFilterPreAnalysis, pushDownVirtualColumnsForLhsExprs);
        if (joinFilterAnalysis.isCanPushDown()) {
            // noinspection OptionalGetWithoutIsPresent isCanPushDown checks isPresent
            leftFilters.add(joinFilterAnalysis.getPushDownFilter().get());
        }
        if (joinFilterAnalysis.isRetainAfterJoin()) {
            rightFilters.add(joinFilterAnalysis.getOriginalFilter());
        }
    }
    return new JoinFilterSplit(Filters.maybeAnd(leftFilters).orElse(null), Filters.maybeAnd(rightFilters).orElse(null), new HashSet<>(pushDownVirtualColumnsForLhsExprs.values()));
}

Example 3 with VirtualColumn

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

the class JoinFilterAnalyzer method rewriteFilterDirect.

private static JoinFilterAnalysis rewriteFilterDirect(Filter filterClause, JoinFilterPreAnalysis joinFilterPreAnalysis, Map<Expr, VirtualColumn> pushDownVirtualColumnsForLhsExprs) {
    if (!filterClause.supportsRequiredColumnRewrite()) {
        return JoinFilterAnalysis.createNoPushdownFilterAnalysis(filterClause);
    }
    List<Filter> newFilters = new ArrayList<>();
    // we only support direct rewrites of filters that reference a single column
    String reqColumn = filterClause.getRequiredColumns().iterator().next();
    List<JoinFilterColumnCorrelationAnalysis> correlationAnalyses = joinFilterPreAnalysis.getCorrelationsByDirectFilteringColumn().get(reqColumn);
    if (correlationAnalyses == null) {
        return JoinFilterAnalysis.createNoPushdownFilterAnalysis(filterClause);
    }
    for (JoinFilterColumnCorrelationAnalysis correlationAnalysis : correlationAnalyses) {
        if (correlationAnalysis.supportsPushDown()) {
            for (String correlatedBaseColumn : correlationAnalysis.getBaseColumns()) {
                Filter rewrittenFilter = filterClause.rewriteRequiredColumns(ImmutableMap.of(reqColumn, correlatedBaseColumn));
                newFilters.add(rewrittenFilter);
            }
            for (Expr correlatedBaseExpr : correlationAnalysis.getBaseExpressions()) {
                // We need to create a virtual column for the expressions when pushing down
                VirtualColumn pushDownVirtualColumn = pushDownVirtualColumnsForLhsExprs.computeIfAbsent(correlatedBaseExpr, (expr) -> {
                    String vcName = getCorrelatedBaseExprVirtualColumnName(pushDownVirtualColumnsForLhsExprs.size());
                    return new ExpressionVirtualColumn(vcName, correlatedBaseExpr, ColumnType.STRING);
                });
                Filter rewrittenFilter = filterClause.rewriteRequiredColumns(ImmutableMap.of(reqColumn, pushDownVirtualColumn.getOutputName()));
                newFilters.add(rewrittenFilter);
            }
        }
    }
    if (newFilters.isEmpty()) {
        return JoinFilterAnalysis.createNoPushdownFilterAnalysis(filterClause);
    }
    return new JoinFilterAnalysis(false, filterClause, Filters.maybeAnd(newFilters).orElse(null));
}

Example 4 with VirtualColumn

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

the class GroupByTimeseriesQueryRunnerTest method constructorFeeder.

@SuppressWarnings("unchecked")
@Parameterized.Parameters(name = "{0}, vectorize = {1}")
public static Iterable<Object[]> constructorFeeder() {
    GroupByQueryConfig config = new GroupByQueryConfig();
    config.setMaxIntermediateRows(10000);
    final Pair<GroupByQueryRunnerFactory, Closer> factoryAndCloser = GroupByQueryRunnerTest.makeQueryRunnerFactory(config);
    final GroupByQueryRunnerFactory factory = factoryAndCloser.lhs;
    RESOURCE_CLOSER.register(factoryAndCloser.rhs);
    final List<Object[]> constructors = new ArrayList<>();
    for (QueryRunner<ResultRow> runner : QueryRunnerTestHelper.makeQueryRunners(factory)) {
        final QueryRunner modifiedRunner = new QueryRunner() {

            @Override
            public Sequence run(QueryPlus queryPlus, ResponseContext responseContext) {
                TimeseriesQuery tsQuery = (TimeseriesQuery) queryPlus.getQuery();
                QueryRunner<ResultRow> newRunner = factory.mergeRunners(Execs.directExecutor(), ImmutableList.of(runner));
                QueryToolChest toolChest = factory.getToolchest();
                newRunner = new FinalizeResultsQueryRunner<>(toolChest.mergeResults(toolChest.preMergeQueryDecoration(newRunner)), toolChest);
                final String timeDimension = tsQuery.getTimestampResultField();
                final List<VirtualColumn> virtualColumns = new ArrayList<>(Arrays.asList(tsQuery.getVirtualColumns().getVirtualColumns()));
                Map<String, Object> theContext = tsQuery.getContext();
                if (timeDimension != null) {
                    theContext = new HashMap<>(tsQuery.getContext());
                    final PeriodGranularity granularity = (PeriodGranularity) tsQuery.getGranularity();
                    virtualColumns.add(new ExpressionVirtualColumn("v0", StringUtils.format("timestamp_floor(__time, '%s')", granularity.getPeriod()), ColumnType.LONG, TestExprMacroTable.INSTANCE));
                    theContext.put(GroupByQuery.CTX_TIMESTAMP_RESULT_FIELD, timeDimension);
                    theContext.put(GroupByQuery.CTX_TIMESTAMP_RESULT_FIELD_GRANULARITY, granularity);
                    theContext.put(GroupByQuery.CTX_TIMESTAMP_RESULT_FIELD_INDEX, 0);
                }
                GroupByQuery newQuery = GroupByQuery.builder().setDataSource(tsQuery.getDataSource()).setQuerySegmentSpec(tsQuery.getQuerySegmentSpec()).setGranularity(tsQuery.getGranularity()).setDimFilter(tsQuery.getDimensionsFilter()).setDimensions(timeDimension == null ? ImmutableList.of() : ImmutableList.of(new DefaultDimensionSpec("v0", timeDimension, ColumnType.LONG))).setAggregatorSpecs(tsQuery.getAggregatorSpecs()).setPostAggregatorSpecs(tsQuery.getPostAggregatorSpecs()).setVirtualColumns(VirtualColumns.create(virtualColumns)).setContext(theContext).build();
                return Sequences.map(newRunner.run(queryPlus.withQuery(newQuery), responseContext), new Function<ResultRow, Result<TimeseriesResultValue>>() {

                    @Override
                    public Result<TimeseriesResultValue> apply(final ResultRow input) {
                        final MapBasedRow mapBasedRow = input.toMapBasedRow(newQuery);
                        return new Result<>(mapBasedRow.getTimestamp(), new TimeseriesResultValue(mapBasedRow.getEvent()));
                    }
                });
            }

            @Override
            public String toString() {
                return runner.toString();
            }
        };
        for (boolean vectorize : ImmutableList.of(false, true)) {
            // Add vectorization tests for any indexes that support it.
            if (!vectorize || QueryRunnerTestHelper.isTestRunnerVectorizable(runner)) {
                constructors.add(new Object[] { modifiedRunner, vectorize });
            }
        }
    }
    return constructors;
}

Example 5 with VirtualColumn

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

the class JoinFilterAnalyzerTest method test_filterPushDown_factToRegionToCountryLeftFilterOnRHSJoinConditionColumnsHelper.

private void test_filterPushDown_factToRegionToCountryLeftFilterOnRHSJoinConditionColumnsHelper(boolean hasLhsExpressionInJoinCondition) {
    Filter expressionFilter = new ExpressionDimFilter("\"rtc.countryIsoCode\" == 'CA'", ExprMacroTable.nil()).toFilter();
    Filter specialSelectorFilter = new SelectorFilter("rtc.countryIsoCode", "CA") {

        @Override
        public boolean supportsRequiredColumnRewrite() {
            return false;
        }
    };
    Filter originalFilter = new AndFilter(ImmutableList.of(new SelectorFilter("r1.regionIsoCode", "ON"), new SelectorFilter("rtc.countryIsoCode", "CA"), specialSelectorFilter, new BoundFilter(new BoundDimFilter("rtc.countryIsoCode", "CA", "CB", false, false, null, null, null)), expressionFilter, new InDimFilter("rtc.countryIsoCode", ImmutableSet.of("CA", "CA2", "CA3"), null, null).toFilter(), new OrFilter(ImmutableList.of(new SelectorFilter("channel", "#fr.wikipedia"), new SelectorFilter("rtc.countryIsoCode", "QQQ"), new BoundFilter(new BoundDimFilter("rtc.countryIsoCode", "YYY", "ZZZ", false, false, null, null, null)))), new OrFilter(ImmutableList.of(new SelectorFilter("namespace", "Main"), new SelectorFilter("rtc.countryIsoCode", "ABCDEF"), new SelectorFilter("rtc.countryName", "Canada"), new BoundFilter(new BoundDimFilter("rtc.countryIsoCode", "XYZXYZ", "XYZXYZ", false, false, null, null, null))))));
    JoinableClause factToRegionClause;
    if (hasLhsExpressionInJoinCondition) {
        factToRegionClause = new JoinableClause(FACT_TO_REGION_PREFIX, new IndexedTableJoinable(regionsTable), JoinType.LEFT, JoinConditionAnalysis.forExpression(StringUtils.format("\"%sregionIsoCode\" == upper(lower(regionIsoCode)) && \"%scountryIsoCode\" == upper(lower(countryIsoCode))", FACT_TO_REGION_PREFIX, FACT_TO_REGION_PREFIX), FACT_TO_REGION_PREFIX, ExprMacroTable.nil()));
    } else {
        factToRegionClause = factToRegion(JoinType.LEFT);
    }
    List<JoinableClause> joinableClauses = ImmutableList.of(factToRegionClause, regionToCountry(JoinType.LEFT));
    JoinFilterPreAnalysis joinFilterPreAnalysis = makeDefaultConfigPreAnalysis(originalFilter, joinableClauses, VirtualColumns.EMPTY);
    HashJoinSegmentStorageAdapter adapter = new HashJoinSegmentStorageAdapter(factSegment.asStorageAdapter(), joinableClauses, joinFilterPreAnalysis);
    String rewrittenCountryIsoCodeColumnName = hasLhsExpressionInJoinCondition ? "JOIN-FILTER-PUSHDOWN-VIRTUAL-COLUMN-1" : "countryIsoCode";
    String rewrittenRegionIsoCodeColumnName = hasLhsExpressionInJoinCondition ? "JOIN-FILTER-PUSHDOWN-VIRTUAL-COLUMN-0" : "regionIsoCode";
    Set<VirtualColumn> expectedVirtualColumns;
    if (hasLhsExpressionInJoinCondition) {
        expectedVirtualColumns = ImmutableSet.of(new ExpressionVirtualColumn(rewrittenRegionIsoCodeColumnName, "(upper [(lower [regionIsoCode])])", ColumnType.STRING, ExprMacroTable.nil()), new ExpressionVirtualColumn(rewrittenCountryIsoCodeColumnName, "(upper [(lower [countryIsoCode])])", ColumnType.STRING, ExprMacroTable.nil()));
    } else {
        expectedVirtualColumns = ImmutableSet.of();
    }
    JoinTestHelper.verifyCursors(adapter.makeCursors(originalFilter, Intervals.ETERNITY, VirtualColumns.EMPTY, Granularities.ALL, false, null), ImmutableList.of("page", FACT_TO_REGION_PREFIX + "regionName", REGION_TO_COUNTRY_PREFIX + "countryName"), ImmutableList.of(new Object[] { "Didier Leclair", "Ontario", "Canada" }));
    JoinFilterSplit expectedFilterSplit = new JoinFilterSplit(new AndFilter(ImmutableList.of(new SelectorFilter(rewrittenRegionIsoCodeColumnName, "ON"), new SelectorFilter(rewrittenCountryIsoCodeColumnName, "CA"), new InDimFilter(rewrittenCountryIsoCodeColumnName, ImmutableSet.of("CA"), null, null).toFilter(), new BoundFilter(new BoundDimFilter(rewrittenCountryIsoCodeColumnName, "CA", "CB", false, false, null, null, null)), new InDimFilter(rewrittenCountryIsoCodeColumnName, ImmutableSet.of("CA", "CA2", "CA3"), null, null).toFilter(), new OrFilter(ImmutableList.of(new SelectorFilter("channel", "#fr.wikipedia"), new SelectorFilter(rewrittenCountryIsoCodeColumnName, "QQQ"), new BoundFilter(new BoundDimFilter(rewrittenCountryIsoCodeColumnName, "YYY", "ZZZ", false, false, null, null, null)))), new OrFilter(ImmutableList.of(new SelectorFilter("namespace", "Main"), new SelectorFilter(rewrittenCountryIsoCodeColumnName, "ABCDEF"), new InDimFilter(rewrittenCountryIsoCodeColumnName, ImmutableSet.of("CA"), null, null).toFilter(), new BoundFilter(new BoundDimFilter(rewrittenCountryIsoCodeColumnName, "XYZXYZ", "XYZXYZ", false, false, null, null, null)))))), new AndFilter(ImmutableList.of(specialSelectorFilter, expressionFilter, new OrFilter(ImmutableList.of(new SelectorFilter("namespace", "Main"), new SelectorFilter("rtc.countryIsoCode", "ABCDEF"), new SelectorFilter("rtc.countryName", "Canada"), new BoundFilter(new BoundDimFilter("rtc.countryIsoCode", "XYZXYZ", "XYZXYZ", false, false, null, null, null)))))), expectedVirtualColumns);
    JoinFilterSplit actualFilterSplit = JoinFilterAnalyzer.splitFilter(joinFilterPreAnalysis);
    Assert.assertEquals(expectedFilterSplit, actualFilterSplit);
}