Search in sources :

Example 1 with DoubleMinAggregatorFactory

use of io.druid.query.aggregation.DoubleMinAggregatorFactory in project druid by druid-io.

the class GroupByRules method translateAggregateCall.

/**
   * Translate an AggregateCall to Druid equivalents.
   *
   * @return translated aggregation, or null if translation failed.
   */
private static Aggregation translateAggregateCall(final PlannerContext plannerContext, final RowSignature sourceRowSignature, final Project project, final AggregateCall call, final DruidOperatorTable operatorTable, final List<Aggregation> existingAggregations, final int aggNumber, final boolean approximateCountDistinct) {
    final List<DimFilter> filters = Lists.newArrayList();
    final List<String> rowOrder = sourceRowSignature.getRowOrder();
    final String name = aggOutputName(aggNumber);
    final SqlKind kind = call.getAggregation().getKind();
    final SqlTypeName outputType = call.getType().getSqlTypeName();
    if (call.filterArg >= 0) {
        // AGG(xxx) FILTER(WHERE yyy)
        if (project == null) {
            // We need some kind of projection to support filtered aggregations.
            return null;
        }
        final RexNode expression = project.getChildExps().get(call.filterArg);
        final DimFilter filter = Expressions.toFilter(operatorTable, plannerContext, sourceRowSignature, expression);
        if (filter == null) {
            return null;
        }
        filters.add(filter);
    }
    if (kind == SqlKind.COUNT && call.getArgList().isEmpty()) {
        // COUNT(*)
        return Aggregation.create(new CountAggregatorFactory(name)).filter(makeFilter(filters, sourceRowSignature));
    } else if (kind == SqlKind.COUNT && call.isDistinct()) {
        // COUNT(DISTINCT x)
        return approximateCountDistinct ? APPROX_COUNT_DISTINCT.toDruidAggregation(name, sourceRowSignature, operatorTable, plannerContext, existingAggregations, project, call, makeFilter(filters, sourceRowSignature)) : null;
    } else if (kind == SqlKind.COUNT || kind == SqlKind.SUM || kind == SqlKind.SUM0 || kind == SqlKind.MIN || kind == SqlKind.MAX || kind == SqlKind.AVG) {
        // Built-in agg, not distinct, not COUNT(*)
        boolean forceCount = false;
        final FieldOrExpression input;
        final int inputField = Iterables.getOnlyElement(call.getArgList());
        final RexNode rexNode = Expressions.fromFieldAccess(sourceRowSignature, project, inputField);
        final FieldOrExpression foe = FieldOrExpression.fromRexNode(operatorTable, plannerContext, rowOrder, rexNode);
        if (foe != null) {
            input = foe;
        } else if (rexNode.getKind() == SqlKind.CASE && ((RexCall) rexNode).getOperands().size() == 3) {
            // Possibly a CASE-style filtered aggregation. Styles supported:
            // A: SUM(CASE WHEN x = 'foo' THEN cnt END) => operands (x = 'foo', cnt, null)
            // B: SUM(CASE WHEN x = 'foo' THEN 1 ELSE 0 END) => operands (x = 'foo', 1, 0)
            // C: COUNT(CASE WHEN x = 'foo' THEN 'dummy' END) => operands (x = 'foo', 'dummy', null)
            // If the null and non-null args are switched, "flip" is set, which negates the filter.
            final RexCall caseCall = (RexCall) rexNode;
            final boolean flip = RexLiteral.isNullLiteral(caseCall.getOperands().get(1)) && !RexLiteral.isNullLiteral(caseCall.getOperands().get(2));
            final RexNode arg1 = caseCall.getOperands().get(flip ? 2 : 1);
            final RexNode arg2 = caseCall.getOperands().get(flip ? 1 : 2);
            // Operand 1: Filter
            final DimFilter filter = Expressions.toFilter(operatorTable, plannerContext, sourceRowSignature, caseCall.getOperands().get(0));
            if (filter == null) {
                return null;
            } else {
                filters.add(flip ? new NotDimFilter(filter) : filter);
            }
            if (call.getAggregation().getKind() == SqlKind.COUNT && arg1 instanceof RexLiteral && !RexLiteral.isNullLiteral(arg1) && RexLiteral.isNullLiteral(arg2)) {
                // Case C
                forceCount = true;
                input = null;
            } else if (call.getAggregation().getKind() == SqlKind.SUM && arg1 instanceof RexLiteral && ((Number) RexLiteral.value(arg1)).intValue() == 1 && arg2 instanceof RexLiteral && ((Number) RexLiteral.value(arg2)).intValue() == 0) {
                // Case B
                forceCount = true;
                input = null;
            } else if (RexLiteral.isNullLiteral(arg2)) {
                // Maybe case A
                input = FieldOrExpression.fromRexNode(operatorTable, plannerContext, rowOrder, arg1);
                if (input == null) {
                    return null;
                }
            } else {
                // Can't translate CASE into a filter.
                return null;
            }
        } else {
            // Can't translate operand.
            return null;
        }
        if (!forceCount) {
            Preconditions.checkNotNull(input, "WTF?! input was null for non-COUNT aggregation");
        }
        if (forceCount || kind == SqlKind.COUNT) {
            // COUNT(x)
            return Aggregation.create(new CountAggregatorFactory(name)).filter(makeFilter(filters, sourceRowSignature));
        } else {
            // Built-in aggregator that is not COUNT.
            final Aggregation retVal;
            final String fieldName = input.getFieldName();
            final String expression = input.getExpression();
            final boolean isLong = SqlTypeName.INT_TYPES.contains(outputType) || SqlTypeName.TIMESTAMP == outputType || SqlTypeName.DATE == outputType;
            if (kind == SqlKind.SUM || kind == SqlKind.SUM0) {
                retVal = isLong ? Aggregation.create(new LongSumAggregatorFactory(name, fieldName, expression)) : Aggregation.create(new DoubleSumAggregatorFactory(name, fieldName, expression));
            } else if (kind == SqlKind.MIN) {
                retVal = isLong ? Aggregation.create(new LongMinAggregatorFactory(name, fieldName, expression)) : Aggregation.create(new DoubleMinAggregatorFactory(name, fieldName, expression));
            } else if (kind == SqlKind.MAX) {
                retVal = isLong ? Aggregation.create(new LongMaxAggregatorFactory(name, fieldName, expression)) : Aggregation.create(new DoubleMaxAggregatorFactory(name, fieldName, expression));
            } else if (kind == SqlKind.AVG) {
                final String sumName = aggInternalName(aggNumber, "sum");
                final String countName = aggInternalName(aggNumber, "count");
                final AggregatorFactory sum = isLong ? new LongSumAggregatorFactory(sumName, fieldName, expression) : new DoubleSumAggregatorFactory(sumName, fieldName, expression);
                final AggregatorFactory count = new CountAggregatorFactory(countName);
                retVal = Aggregation.create(ImmutableList.of(sum, count), new ArithmeticPostAggregator(name, "quotient", ImmutableList.<PostAggregator>of(new FieldAccessPostAggregator(null, sumName), new FieldAccessPostAggregator(null, countName))));
            } else {
                // Not reached.
                throw new ISE("WTF?! Kind[%s] got into the built-in aggregator path somehow?!", kind);
            }
            return retVal.filter(makeFilter(filters, sourceRowSignature));
        }
    } else {
        // Not a built-in aggregator, check operator table.
        final SqlAggregator sqlAggregator = operatorTable.lookupAggregator(call.getAggregation().getName());
        return sqlAggregator != null ? sqlAggregator.toDruidAggregation(name, sourceRowSignature, operatorTable, plannerContext, existingAggregations, project, call, makeFilter(filters, sourceRowSignature)) : null;
    }
}
Also used : RexLiteral(org.apache.calcite.rex.RexLiteral) ArithmeticPostAggregator(io.druid.query.aggregation.post.ArithmeticPostAggregator) DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) SqlTypeName(org.apache.calcite.sql.type.SqlTypeName) LongSumAggregatorFactory(io.druid.query.aggregation.LongSumAggregatorFactory) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) LongMinAggregatorFactory(io.druid.query.aggregation.LongMinAggregatorFactory) RexCall(org.apache.calcite.rex.RexCall) Aggregation(io.druid.sql.calcite.aggregation.Aggregation) ISE(io.druid.java.util.common.ISE) LongMaxAggregatorFactory(io.druid.query.aggregation.LongMaxAggregatorFactory) NotDimFilter(io.druid.query.filter.NotDimFilter) FieldAccessPostAggregator(io.druid.query.aggregation.post.FieldAccessPostAggregator) DoubleSumAggregatorFactory(io.druid.query.aggregation.DoubleSumAggregatorFactory) PostAggregator(io.druid.query.aggregation.PostAggregator) FieldAccessPostAggregator(io.druid.query.aggregation.post.FieldAccessPostAggregator) ArithmeticPostAggregator(io.druid.query.aggregation.post.ArithmeticPostAggregator) SqlKind(org.apache.calcite.sql.SqlKind) CountAggregatorFactory(io.druid.query.aggregation.CountAggregatorFactory) DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) LongMaxAggregatorFactory(io.druid.query.aggregation.LongMaxAggregatorFactory) DoubleSumAggregatorFactory(io.druid.query.aggregation.DoubleSumAggregatorFactory) AggregatorFactory(io.druid.query.aggregation.AggregatorFactory) LongMinAggregatorFactory(io.druid.query.aggregation.LongMinAggregatorFactory) PostAggregatorFactory(io.druid.sql.calcite.aggregation.PostAggregatorFactory) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) LongSumAggregatorFactory(io.druid.query.aggregation.LongSumAggregatorFactory) CountAggregatorFactory(io.druid.query.aggregation.CountAggregatorFactory) SqlAggregator(io.druid.sql.calcite.aggregation.SqlAggregator) ApproxCountDistinctSqlAggregator(io.druid.sql.calcite.aggregation.ApproxCountDistinctSqlAggregator) DimFilter(io.druid.query.filter.DimFilter) NotDimFilter(io.druid.query.filter.NotDimFilter) AndDimFilter(io.druid.query.filter.AndDimFilter) RexNode(org.apache.calcite.rex.RexNode)

Example 2 with DoubleMinAggregatorFactory

use of io.druid.query.aggregation.DoubleMinAggregatorFactory in project druid by druid-io.

the class GroupByTimeseriesQueryRunnerTest method testFullOnTimeseriesMaxMin.

// GroupBy handles timestamps differently when granularity is ALL
@Test
public void testFullOnTimeseriesMaxMin() {
    TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource(QueryRunnerTestHelper.dataSource).granularity(Granularities.ALL).intervals(QueryRunnerTestHelper.fullOnInterval).aggregators(Arrays.asList(new DoubleMaxAggregatorFactory("maxIndex", "index"), new DoubleMinAggregatorFactory("minIndex", "index"))).descending(descending).build();
    DateTime expectedEarliest = new DateTime("1970-01-01");
    DateTime expectedLast = new DateTime("2011-04-15");
    Iterable<Result<TimeseriesResultValue>> results = Sequences.toList(runner.run(query, CONTEXT), Lists.<Result<TimeseriesResultValue>>newArrayList());
    Result<TimeseriesResultValue> result = results.iterator().next();
    Assert.assertEquals(expectedEarliest, result.getTimestamp());
    Assert.assertFalse(String.format("Timestamp[%s] > expectedLast[%s]", result.getTimestamp(), expectedLast), result.getTimestamp().isAfter(expectedLast));
    final TimeseriesResultValue value = result.getValue();
    Assert.assertEquals(result.toString(), 1870.06103515625, value.getDoubleMetric("maxIndex"), 0.0);
    Assert.assertEquals(result.toString(), 59.02102279663086, value.getDoubleMetric("minIndex"), 0.0);
}
Also used : TimeseriesResultValue(io.druid.query.timeseries.TimeseriesResultValue) DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) TimeseriesQuery(io.druid.query.timeseries.TimeseriesQuery) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) DateTime(org.joda.time.DateTime) Result(io.druid.query.Result) TimeseriesQueryRunnerTest(io.druid.query.timeseries.TimeseriesQueryRunnerTest) Test(org.junit.Test)

Example 3 with DoubleMinAggregatorFactory

use of io.druid.query.aggregation.DoubleMinAggregatorFactory in project druid by druid-io.

the class TopNQueryRunnerTest method testTopNOverNullDimensionWithFilter.

@Test
public void testTopNOverNullDimensionWithFilter() {
    TopNQuery query = new TopNQueryBuilder().dataSource(QueryRunnerTestHelper.dataSource).granularity(QueryRunnerTestHelper.allGran).dimension("null_column").filters(new SelectorDimFilter("null_column", null, null)).metric(QueryRunnerTestHelper.indexMetric).threshold(4).intervals(QueryRunnerTestHelper.fullOnInterval).aggregators(Lists.newArrayList(Iterables.concat(QueryRunnerTestHelper.commonAggregators, Lists.newArrayList(new DoubleMaxAggregatorFactory("maxIndex", "index"), new DoubleMinAggregatorFactory("minIndex", "index"))))).postAggregators(Arrays.<PostAggregator>asList(QueryRunnerTestHelper.addRowsIndexConstant)).build();
    Map<String, Object> map = Maps.newHashMap();
    map.put("null_column", null);
    map.put("rows", 1209L);
    map.put("index", 503332.5071372986D);
    map.put("addRowsIndexConstant", 504542.5071372986D);
    map.put("uniques", QueryRunnerTestHelper.UNIQUES_9);
    map.put("maxIndex", 1870.06103515625D);
    map.put("minIndex", 59.02102279663086D);
    List<Result<TopNResultValue>> expectedResults = Arrays.asList(new Result<>(new DateTime("2011-01-12T00:00:00.000Z"), new TopNResultValue(Arrays.asList(map))));
    assertExpectedResults(expectedResults, query);
}
Also used : DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) HyperUniqueFinalizingPostAggregator(io.druid.query.aggregation.hyperloglog.HyperUniqueFinalizingPostAggregator) PostAggregator(io.druid.query.aggregation.PostAggregator) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) DateTime(org.joda.time.DateTime) Result(io.druid.query.Result) SelectorDimFilter(io.druid.query.filter.SelectorDimFilter) Test(org.junit.Test)

Example 4 with DoubleMinAggregatorFactory

use of io.druid.query.aggregation.DoubleMinAggregatorFactory in project druid by druid-io.

the class TopNQueryRunnerTest method testFullOnTopNDimExtractionAllNulls.

@Test
public void testFullOnTopNDimExtractionAllNulls() {
    String jsFn = "function(str) { return null; }";
    ExtractionFn jsExtractionFn = new JavaScriptExtractionFn(jsFn, false, JavaScriptConfig.getEnabledInstance());
    TopNQuery query = new TopNQueryBuilder().dataSource(QueryRunnerTestHelper.dataSource).granularity(QueryRunnerTestHelper.allGran).dimension(new ExtractionDimensionSpec(QueryRunnerTestHelper.marketDimension, QueryRunnerTestHelper.marketDimension, jsExtractionFn)).metric(QueryRunnerTestHelper.indexMetric).threshold(4).intervals(QueryRunnerTestHelper.fullOnInterval).aggregators(Lists.<AggregatorFactory>newArrayList(Iterables.concat(QueryRunnerTestHelper.commonAggregators, Lists.newArrayList(new DoubleMaxAggregatorFactory("maxIndex", "index"), new DoubleMinAggregatorFactory("minIndex", "index"))))).postAggregators(Arrays.<PostAggregator>asList(QueryRunnerTestHelper.addRowsIndexConstant)).build();
    Map<String, Object> expectedMap = new HashMap<>();
    expectedMap.put(QueryRunnerTestHelper.marketDimension, null);
    expectedMap.put("rows", 1209L);
    expectedMap.put("index", 503332.5071372986D);
    expectedMap.put("addRowsIndexConstant", 504542.5071372986D);
    expectedMap.put("uniques", 9.019833517963864);
    expectedMap.put("maxIndex", 1870.06103515625D);
    expectedMap.put("minIndex", 59.02102279663086D);
    List<Result<TopNResultValue>> expectedResults = Arrays.asList(new Result<TopNResultValue>(new DateTime("2011-01-12T00:00:00.000Z"), new TopNResultValue(Arrays.<Map<String, Object>>asList(expectedMap))));
    assertExpectedResults(expectedResults, query);
}
Also used : DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) HyperUniqueFinalizingPostAggregator(io.druid.query.aggregation.hyperloglog.HyperUniqueFinalizingPostAggregator) PostAggregator(io.druid.query.aggregation.PostAggregator) HashMap(java.util.HashMap) LinkedHashMap(java.util.LinkedHashMap) JavaScriptExtractionFn(io.druid.query.extraction.JavaScriptExtractionFn) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) DateTime(org.joda.time.DateTime) Result(io.druid.query.Result) LookupExtractionFn(io.druid.query.lookup.LookupExtractionFn) ExtractionFn(io.druid.query.extraction.ExtractionFn) RegexDimExtractionFn(io.druid.query.extraction.RegexDimExtractionFn) DimExtractionFn(io.druid.query.extraction.DimExtractionFn) StrlenExtractionFn(io.druid.query.extraction.StrlenExtractionFn) JavaScriptExtractionFn(io.druid.query.extraction.JavaScriptExtractionFn) TimeFormatExtractionFn(io.druid.query.extraction.TimeFormatExtractionFn) ExtractionDimensionSpec(io.druid.query.dimension.ExtractionDimensionSpec) Test(org.junit.Test)

Example 5 with DoubleMinAggregatorFactory

use of io.druid.query.aggregation.DoubleMinAggregatorFactory in project druid by druid-io.

the class TopNQueryRunnerTest method testFullOnTopNLongColumn.

@Test
public void testFullOnTopNLongColumn() {
    TopNQuery query = new TopNQueryBuilder().dataSource(QueryRunnerTestHelper.dataSource).granularity(QueryRunnerTestHelper.allGran).dimension(new DefaultDimensionSpec("qualityLong", "ql_alias", ValueType.LONG)).metric("maxIndex").threshold(4).intervals(QueryRunnerTestHelper.fullOnInterval).aggregators(Lists.<AggregatorFactory>newArrayList(Iterables.concat(QueryRunnerTestHelper.commonAggregators, Lists.newArrayList(new DoubleMaxAggregatorFactory("maxIndex", "index"), new DoubleMinAggregatorFactory("minIndex", "index"))))).postAggregators(Arrays.<PostAggregator>asList(QueryRunnerTestHelper.addRowsIndexConstant)).build();
    List<Result<TopNResultValue>> expectedResults = Arrays.asList(new Result<TopNResultValue>(new DateTime("2011-01-12T00:00:00.000Z"), new TopNResultValue(Arrays.<Map<String, Object>>asList(ImmutableMap.<String, Object>builder().put("ql_alias", 1400L).put(QueryRunnerTestHelper.indexMetric, 217725.42022705078D).put("rows", 279L).put("addRowsIndexConstant", 218005.42022705078D).put("uniques", QueryRunnerTestHelper.UNIQUES_1).put("maxIndex", 1870.06103515625D).put("minIndex", 91.27055358886719D).build(), ImmutableMap.<String, Object>builder().put("ql_alias", 1600L).put(QueryRunnerTestHelper.indexMetric, 210865.67966461182D).put("rows", 279L).put("addRowsIndexConstant", 211145.67966461182D).put("uniques", QueryRunnerTestHelper.UNIQUES_1).put("maxIndex", 1862.7379150390625D).put("minIndex", 99.2845230102539D).build(), ImmutableMap.<String, Object>builder().put("ql_alias", 1000L).put(QueryRunnerTestHelper.indexMetric, 12270.807106018066D).put("rows", 93L).put("addRowsIndexConstant", 12364.807106018066D).put("uniques", QueryRunnerTestHelper.UNIQUES_1).put("maxIndex", 277.2735290527344D).put("minIndex", 71.31593322753906D).build(), ImmutableMap.<String, Object>builder().put("ql_alias", 1200L).put(QueryRunnerTestHelper.indexMetric, 12086.472755432129D).put("rows", 93L).put("addRowsIndexConstant", 12180.472755432129D).put("uniques", QueryRunnerTestHelper.UNIQUES_1).put("maxIndex", 193.78756713867188D).put("minIndex", 84.71052551269531D).build()))));
    assertExpectedResults(expectedResults, query);
}
Also used : DoubleMaxAggregatorFactory(io.druid.query.aggregation.DoubleMaxAggregatorFactory) HyperUniqueFinalizingPostAggregator(io.druid.query.aggregation.hyperloglog.HyperUniqueFinalizingPostAggregator) PostAggregator(io.druid.query.aggregation.PostAggregator) DoubleMinAggregatorFactory(io.druid.query.aggregation.DoubleMinAggregatorFactory) DefaultDimensionSpec(io.druid.query.dimension.DefaultDimensionSpec) DateTime(org.joda.time.DateTime) Result(io.druid.query.Result) Test(org.junit.Test)

Aggregations

DoubleMinAggregatorFactory (io.druid.query.aggregation.DoubleMinAggregatorFactory)36 DoubleMaxAggregatorFactory (io.druid.query.aggregation.DoubleMaxAggregatorFactory)33 Result (io.druid.query.Result)30 Test (org.junit.Test)27 PostAggregator (io.druid.query.aggregation.PostAggregator)26 DateTime (org.joda.time.DateTime)26 HyperUniqueFinalizingPostAggregator (io.druid.query.aggregation.hyperloglog.HyperUniqueFinalizingPostAggregator)15 HashMap (java.util.HashMap)11 DefaultDimensionSpec (io.druid.query.dimension.DefaultDimensionSpec)10 LinkedHashMap (java.util.LinkedHashMap)10 ArithmeticPostAggregator (io.druid.query.aggregation.post.ArithmeticPostAggregator)8 FieldAccessPostAggregator (io.druid.query.aggregation.post.FieldAccessPostAggregator)8 LookupExtractionFn (io.druid.query.lookup.LookupExtractionFn)8 ConstantPostAggregator (io.druid.query.aggregation.post.ConstantPostAggregator)7 ExtractionDimensionSpec (io.druid.query.dimension.ExtractionDimensionSpec)7 TopNQueryBuilder (io.druid.query.topn.TopNQueryBuilder)7 DoubleSumAggregatorFactory (io.druid.query.aggregation.DoubleSumAggregatorFactory)6 LongSumAggregatorFactory (io.druid.query.aggregation.LongSumAggregatorFactory)6 FilteredAggregatorFactory (io.druid.query.aggregation.FilteredAggregatorFactory)5 HyperUniquesAggregatorFactory (io.druid.query.aggregation.hyperloglog.HyperUniquesAggregatorFactory)5