Examples with OrDimFilter org.apache.druid.query.filter.OrDimFilter used on opensource projects

Example 16 with OrDimFilter

use of org.apache.druid.query.filter.OrDimFilter in project druid by druid-io.

the class FilteredAggregatorTest method testAggregateWithOrFilter.

@Test
public void testAggregateWithOrFilter() {
    final float[] values = { 0.15f, 0.27f, 0.14f };
    final TestFloatColumnSelector selector = new TestFloatColumnSelector(values);
    FilteredAggregatorFactory factory = new FilteredAggregatorFactory(new DoubleSumAggregatorFactory("billy", "value"), new OrDimFilter(Lists.newArrayList(new SelectorDimFilter("dim", "a", null), new SelectorDimFilter("dim", "b", null))));
    FilteredAggregator agg = (FilteredAggregator) factory.factorize(makeColumnSelector(selector));
    double expectedFirst = new Float(values[0]).doubleValue();
    double expectedSecond = new Float(values[1]).doubleValue() + expectedFirst;
    double expectedThird = expectedSecond + new Float(values[2]).doubleValue();
    assertValues(agg, selector, expectedFirst, expectedSecond, expectedThird);
}

Example 17 with OrDimFilter

use of org.apache.druid.query.filter.OrDimFilter in project druid by druid-io.

the class CombineAndSimplifyBounds method doSimplify.

/**
 * Simplify BoundDimFilters that are children of an OR or an AND.
 *
 * @param children    the filters
 * @param disjunction true for OR, false for AND
 *
 * @return simplified filters
 */
private static DimFilter doSimplify(final List<DimFilter> children, boolean disjunction) {
    // Copy the list of child filters. We'll modify the copy and eventually return it.
    final List<DimFilter> newChildren = Lists.newArrayList(children);
    // Group Bound filters by dimension, extractionFn, and comparator and compute a RangeSet for each one.
    final Map<BoundRefKey, List<BoundDimFilter>> bounds = new HashMap<>();
    // all and/or filters have at least 1 child
    boolean allFalse = true;
    for (final DimFilter child : newChildren) {
        if (child instanceof BoundDimFilter) {
            final BoundDimFilter bound = (BoundDimFilter) child;
            final BoundRefKey boundRefKey = BoundRefKey.from(bound);
            final List<BoundDimFilter> filterList = bounds.computeIfAbsent(boundRefKey, k -> new ArrayList<>());
            filterList.add(bound);
            allFalse = false;
        } else {
            allFalse &= child instanceof FalseDimFilter;
        }
    }
    // short circuit if can never be true
    if (allFalse) {
        return Filtration.matchNothing();
    }
    // Try to simplify filters within each group.
    for (Map.Entry<BoundRefKey, List<BoundDimFilter>> entry : bounds.entrySet()) {
        final BoundRefKey boundRefKey = entry.getKey();
        final List<BoundDimFilter> filterList = entry.getValue();
        // Create a RangeSet for this group.
        final RangeSet<BoundValue> rangeSet = disjunction ? RangeSets.unionRanges(Bounds.toRanges(filterList)) : RangeSets.intersectRanges(Bounds.toRanges(filterList));
        if (rangeSet.asRanges().size() < filterList.size()) {
            // We found a simplification. Remove the old filters and add new ones.
            for (final BoundDimFilter bound : filterList) {
                if (!newChildren.remove(bound)) {
                    // Don't expect this to happen, but include it as a sanity check.
                    throw new ISE("Tried to remove bound, but couldn't");
                }
            }
            if (rangeSet.asRanges().isEmpty()) {
                // range set matches nothing, equivalent to FALSE
                newChildren.add(Filtration.matchNothing());
            }
            for (final Range<BoundValue> range : rangeSet.asRanges()) {
                if (!range.hasLowerBound() && !range.hasUpperBound()) {
                    // range matches all, equivalent to TRUE
                    newChildren.add(Filtration.matchEverything());
                } else {
                    newChildren.add(Bounds.toFilter(boundRefKey, range));
                }
            }
        }
    }
    // Finally: Go through newChildren, removing or potentially exiting early based on TRUE / FALSE marker filters.
    Preconditions.checkState(newChildren.size() > 0, "newChildren.size > 0");
    final Iterator<DimFilter> iterator = newChildren.iterator();
    while (iterator.hasNext()) {
        final DimFilter newChild = iterator.next();
        if (Filtration.matchNothing().equals(newChild)) {
            // AND with FALSE => always false, short circuit
            if (disjunction) {
                iterator.remove();
            } else {
                return Filtration.matchNothing();
            }
        } else if (Filtration.matchEverything().equals(newChild)) {
            // AND with TRUE => ignore
            if (disjunction) {
                return Filtration.matchEverything();
            } else {
                iterator.remove();
            }
        }
    }
    if (newChildren.isEmpty()) {
        // If "newChildren" is empty at this point, it must have consisted entirely of TRUE / FALSE marker filters.
        if (disjunction) {
            // Must have been all FALSE filters (the only kind we would have removed above).
            return Filtration.matchNothing();
        } else {
            // Must have been all TRUE filters (the only kind we would have removed above).
            return Filtration.matchEverything();
        }
    } else if (newChildren.size() == 1) {
        return newChildren.get(0);
    } else {
        return disjunction ? new OrDimFilter(newChildren) : new AndDimFilter(newChildren);
    }
}

Example 18 with OrDimFilter

use of org.apache.druid.query.filter.OrDimFilter in project druid by druid-io.

the class BottomUpTransform method apply0.

private DimFilter apply0(final DimFilter filter) {
    if (filter instanceof AndDimFilter) {
        final List<DimFilter> oldFilters = ((AndDimFilter) filter).getFields();
        final List<DimFilter> newFilters = new ArrayList<>();
        for (DimFilter oldFilter : oldFilters) {
            final DimFilter newFilter = apply0(oldFilter);
            if (newFilter != null) {
                newFilters.add(newFilter);
            }
        }
        if (!newFilters.equals(oldFilters)) {
            return checkedProcess(new AndDimFilter(newFilters));
        } else {
            return checkedProcess(filter);
        }
    } else if (filter instanceof OrDimFilter) {
        final List<DimFilter> oldFilters = ((OrDimFilter) filter).getFields();
        final List<DimFilter> newFilters = new ArrayList<>();
        for (DimFilter oldFilter : oldFilters) {
            final DimFilter newFilter = apply0(oldFilter);
            if (newFilter != null) {
                newFilters.add(newFilter);
            }
        }
        if (!newFilters.equals(oldFilters)) {
            return checkedProcess(new OrDimFilter(newFilters));
        } else {
            return checkedProcess(filter);
        }
    } else if (filter instanceof NotDimFilter) {
        final DimFilter oldFilter = ((NotDimFilter) filter).getField();
        final DimFilter newFilter = apply0(oldFilter);
        if (!oldFilter.equals(newFilter)) {
            return checkedProcess(new NotDimFilter(newFilter));
        } else {
            return checkedProcess(filter);
        }
    } else {
        return checkedProcess(filter);
    }
}

Example 19 with OrDimFilter

use of org.apache.druid.query.filter.OrDimFilter in project druid by druid-io.

the class BloomDimFilterSqlTest method testBloomFilters.

@Test
public void testBloomFilters() throws Exception {
    BloomKFilter filter = new BloomKFilter(1500);
    filter.addString("def");
    BloomKFilter filter2 = new BloomKFilter(1500);
    filter.addString("abc");
    byte[] bytes = BloomFilterSerializersModule.bloomKFilterToBytes(filter);
    byte[] bytes2 = BloomFilterSerializersModule.bloomKFilterToBytes(filter2);
    String base64 = StringUtils.encodeBase64String(bytes);
    String base642 = StringUtils.encodeBase64String(bytes2);
    testQuery(StringUtils.format("SELECT COUNT(*) FROM druid.foo WHERE bloom_filter_test(dim1, '%s') OR bloom_filter_test(dim2, '%s')", base64, base642), ImmutableList.of(Druids.newTimeseriesQueryBuilder().dataSource(CalciteTests.DATASOURCE1).intervals(querySegmentSpec(Filtration.eternity())).granularity(Granularities.ALL).filters(new OrDimFilter(new BloomDimFilter("dim1", BloomKFilterHolder.fromBloomKFilter(filter), null), new BloomDimFilter("dim2", BloomKFilterHolder.fromBloomKFilter(filter2), null))).aggregators(aggregators(new CountAggregatorFactory("a0"))).context(QUERY_CONTEXT_DEFAULT).build()), ImmutableList.of(new Object[] { 2L }));
}

Example 20 with OrDimFilter

use of org.apache.druid.query.filter.OrDimFilter in project druid by druid-io.

the class FilteredAggregatorBenchmark method setup.

/**
 * Setup everything common for benchmarking both the incremental-index and the queriable-index.
 */
@Setup
public void setup() {
    log.info("SETUP CALLED AT " + System.currentTimeMillis());
    ComplexMetrics.registerSerde("hyperUnique", new HyperUniquesSerde());
    schemaInfo = GeneratorBasicSchemas.SCHEMA_MAP.get(schema);
    generator = new DataGenerator(schemaInfo.getColumnSchemas(), RNG_SEED, schemaInfo.getDataInterval(), rowsPerSegment);
    filter = new OrDimFilter(Arrays.asList(new BoundDimFilter("dimSequential", "-1", "-1", true, true, null, null, StringComparators.ALPHANUMERIC), new RegexDimFilter("dimSequential", "X", null), new SearchQueryDimFilter("dimSequential", new ContainsSearchQuerySpec("X", false), null), new InDimFilter("dimSequential", Collections.singletonList("X"), null)));
    filteredMetric = new FilteredAggregatorFactory(new CountAggregatorFactory("rows"), filter);
    factory = new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(), new TimeseriesQueryEngine(), QueryBenchmarkUtil.NOOP_QUERYWATCHER);
    GeneratorSchemaInfo basicSchema = GeneratorBasicSchemas.SCHEMA_MAP.get("basic");
    QuerySegmentSpec intervalSpec = new MultipleIntervalSegmentSpec(Collections.singletonList(basicSchema.getDataInterval()));
    List<AggregatorFactory> queryAggs = Collections.singletonList(filteredMetric);
    query = Druids.newTimeseriesQueryBuilder().dataSource("blah").granularity(Granularities.ALL).intervals(intervalSpec).aggregators(queryAggs).descending(descending).build();
}