Examples with Range io.trino.spi.predicate.Range used on opensource projects

Example 1 with Range

use of io.trino.spi.predicate.Range in project trino by trinodb.

the class PushPredicateThroughProjectIntoWindow method extractUpperBound.

private static OptionalInt extractUpperBound(TupleDomain<Symbol> tupleDomain, Symbol symbol) {
    if (tupleDomain.isNone()) {
        return OptionalInt.empty();
    }
    Domain rankingDomain = tupleDomain.getDomains().get().get(symbol);
    if (rankingDomain == null) {
        return OptionalInt.empty();
    }
    ValueSet values = rankingDomain.getValues();
    if (values.isAll() || values.isNone() || values.getRanges().getRangeCount() <= 0) {
        return OptionalInt.empty();
    }
    Range span = values.getRanges().getSpan();
    if (span.isHighUnbounded()) {
        return OptionalInt.empty();
    }
    long upperBound = (Long) span.getHighBoundedValue();
    if (!span.isHighInclusive()) {
        upperBound--;
    }
    if (upperBound >= Integer.MIN_VALUE && upperBound <= Integer.MAX_VALUE) {
        return OptionalInt.of(toIntExact(upperBound));
    }
    return OptionalInt.empty();
}

Example 2 with Range

use of io.trino.spi.predicate.Range in project trino by trinodb.

the class IonSqlQueryBuilder method toPredicate.

private String toPredicate(Domain domain, Type type, int position) {
    checkArgument(domain.getType().isOrderable(), "Domain type must be orderable");
    if (domain.getValues().isNone()) {
        if (domain.isNullAllowed()) {
            return format("s._%d", position + 1) + " = '' ";
        }
        return "FALSE";
    }
    if (domain.getValues().isAll()) {
        if (domain.isNullAllowed()) {
            return "TRUE";
        }
        return format("s._%d", position + 1) + " <> '' ";
    }
    List<String> disjuncts = new ArrayList<>();
    List<Object> singleValues = new ArrayList<>();
    for (Range range : domain.getValues().getRanges().getOrderedRanges()) {
        checkState(!range.isAll());
        if (range.isSingleValue()) {
            singleValues.add(range.getSingleValue());
            continue;
        }
        List<String> rangeConjuncts = new ArrayList<>();
        if (!range.isLowUnbounded()) {
            rangeConjuncts.add(toPredicate(range.isLowInclusive() ? ">=" : ">", range.getLowBoundedValue(), type, position));
        }
        if (!range.isHighUnbounded()) {
            rangeConjuncts.add(toPredicate(range.isHighInclusive() ? "<=" : "<", range.getHighBoundedValue(), type, position));
        }
        // If rangeConjuncts is null, then the range was ALL, which should already have been checked for
        checkState(!rangeConjuncts.isEmpty());
        disjuncts.add("(" + Joiner.on(" AND ").join(rangeConjuncts) + ")");
    }
    // Add back all of the possible single values either as an equality or an IN predicate
    if (singleValues.size() == 1) {
        disjuncts.add(toPredicate("=", getOnlyElement(singleValues), type, position));
    } else if (singleValues.size() > 1) {
        List<String> values = new ArrayList<>();
        for (Object value : singleValues) {
            checkType(type);
            values.add(valueToQuery(type, value));
        }
        disjuncts.add(createColumn(type, position) + " IN (" + Joiner.on(",").join(values) + ")");
    }
    // Add nullability disjuncts
    checkState(!disjuncts.isEmpty());
    if (domain.isNullAllowed()) {
        disjuncts.add(format("s._%d", position + 1) + " = '' ");
    }
    return "(" + Joiner.on(" OR ").join(disjuncts) + ")";
}

Example 3 with Range

use of io.trino.spi.predicate.Range in project trino by trinodb.

the class CassandraClusteringPredicatesExtractor method getClusteringKeysSet.

private static ClusteringPushDownResult getClusteringKeysSet(List<CassandraColumnHandle> clusteringColumns, TupleDomain<ColumnHandle> predicates, VersionNumber cassandraVersion) {
    ImmutableSet.Builder<ColumnHandle> fullyPushedColumnPredicates = ImmutableSet.builder();
    ImmutableList.Builder<String> clusteringColumnSql = ImmutableList.builder();
    int allProcessedClusteringColumns = 0;
    for (CassandraColumnHandle columnHandle : clusteringColumns) {
        Domain domain = predicates.getDomains().get().get(columnHandle);
        if (domain == null) {
            break;
        }
        if (domain.isNullAllowed()) {
            break;
        }
        int currentlyProcessedClusteringColumn = allProcessedClusteringColumns;
        String predicateString = domain.getValues().getValuesProcessor().transform(ranges -> {
            if (ranges.getRangeCount() == 1) {
                fullyPushedColumnPredicates.add(columnHandle);
                return translateRangeIntoCql(columnHandle, getOnlyElement(ranges.getOrderedRanges()));
            }
            if (ranges.getOrderedRanges().stream().allMatch(Range::isSingleValue)) {
                if (isInExpressionNotAllowed(clusteringColumns, cassandraVersion, currentlyProcessedClusteringColumn)) {
                    return translateRangeIntoCql(columnHandle, ranges.getSpan());
                }
                String inValues = ranges.getOrderedRanges().stream().map(range -> toCqlLiteral(columnHandle, range.getSingleValue())).collect(joining(","));
                fullyPushedColumnPredicates.add(columnHandle);
                return CassandraCqlUtils.validColumnName(columnHandle.getName()) + " IN (" + inValues + ")";
            }
            return translateRangeIntoCql(columnHandle, ranges.getSpan());
        }, discreteValues -> {
            if (discreteValues.isInclusive()) {
                if (discreteValues.getValuesCount() == 0) {
                    return null;
                }
                if (discreteValues.getValuesCount() == 1) {
                    fullyPushedColumnPredicates.add(columnHandle);
                    return format("%s = %s", CassandraCqlUtils.validColumnName(columnHandle.getName()), toCqlLiteral(columnHandle, getOnlyElement(discreteValues.getValues())));
                }
                if (isInExpressionNotAllowed(clusteringColumns, cassandraVersion, currentlyProcessedClusteringColumn)) {
                    return null;
                }
                String inValues = discreteValues.getValues().stream().map(columnHandle.getCassandraType()::toCqlLiteral).collect(joining(","));
                fullyPushedColumnPredicates.add(columnHandle);
                return CassandraCqlUtils.validColumnName(columnHandle.getName()) + " IN (" + inValues + " )";
            }
            return null;
        }, allOrNone -> null);
        if (predicateString == null) {
            break;
        }
        clusteringColumnSql.add(predicateString);
        // Check for last clustering column should only be restricted by range condition
        if (predicateString.contains(">") || predicateString.contains("<")) {
            break;
        }
        allProcessedClusteringColumns++;
    }
    List<String> clusteringColumnPredicates = clusteringColumnSql.build();
    return new ClusteringPushDownResult(fullyPushedColumnPredicates.build(), Joiner.on(" AND ").join(clusteringColumnPredicates));
}

Example 4 with Range

use of io.trino.spi.predicate.Range in project trino by trinodb.

the class CassandraPartitionManager method getPartitions.

public CassandraPartitionResult getPartitions(CassandraTableHandle cassandraTableHandle, TupleDomain<ColumnHandle> tupleDomain) {
    // TODO support repeated applyFilter
    checkArgument(cassandraTableHandle.getPartitions().isEmpty(), "getPartitions() currently does not take into account table handle's partitions");
    CassandraTable table = cassandraSession.getTable(cassandraTableHandle.getSchemaTableName());
    // fetch the partitions
    List<CassandraPartition> allPartitions = getCassandraPartitions(table, tupleDomain);
    log.debug("%s.%s #partitions: %d", cassandraTableHandle.getSchemaName(), cassandraTableHandle.getTableName(), allPartitions.size());
    // do a final pass to filter based on fields that could not be used to build the prefix
    List<CassandraPartition> partitions = allPartitions.stream().filter(partition -> tupleDomain.overlaps(partition.getTupleDomain())).collect(toList());
    // All partition key domains will be fully evaluated, so we don't need to include those
    TupleDomain<ColumnHandle> remainingTupleDomain = TupleDomain.none();
    if (!tupleDomain.isNone()) {
        if (partitions.size() == 1 && partitions.get(0).isUnpartitioned()) {
            remainingTupleDomain = tupleDomain;
        } else {
            Set<ColumnHandle> usedPartitionColumns = partitions.stream().flatMap(partition -> Optional.ofNullable(partition.getTupleDomain()).flatMap(partitionTupleDomain -> partitionTupleDomain.getDomains().map(Map::keySet).map(Set::stream)).orElse(Stream.empty())).collect(toImmutableSet());
            remainingTupleDomain = tupleDomain.filter((column, domain) -> !usedPartitionColumns.contains(column));
        }
    }
    // push down indexed column fixed value predicates only for unpartitioned partition which uses token range query
    if ((partitions.size() == 1) && partitions.get(0).isUnpartitioned()) {
        Map<ColumnHandle, Domain> domains = tupleDomain.getDomains().get();
        List<ColumnHandle> indexedColumns = new ArrayList<>();
        // compose partitionId by using indexed column
        StringBuilder sb = new StringBuilder();
        for (Map.Entry<ColumnHandle, Domain> entry : domains.entrySet()) {
            CassandraColumnHandle column = (CassandraColumnHandle) entry.getKey();
            Domain domain = entry.getValue();
            if (column.isIndexed() && domain.isSingleValue()) {
                sb.append(CassandraCqlUtils.validColumnName(column.getName())).append(" = ").append(column.getCassandraType().toCqlLiteral(entry.getValue().getSingleValue()));
                indexedColumns.add(column);
                // Only one indexed column predicate can be pushed down.
                break;
            }
        }
        if (sb.length() > 0) {
            CassandraPartition partition = partitions.get(0);
            TupleDomain<ColumnHandle> filterIndexedColumn = TupleDomain.withColumnDomains(Maps.filterKeys(remainingTupleDomain.getDomains().get(), not(in(indexedColumns))));
            partitions = new ArrayList<>();
            partitions.add(new CassandraPartition(partition.getKey(), sb.toString(), filterIndexedColumn, true));
            return new CassandraPartitionResult(partitions, filterIndexedColumn);
        }
    }
    return new CassandraPartitionResult(partitions, remainingTupleDomain);
}

Example 5 with Range

use of io.trino.spi.predicate.Range in project trino by trinodb.

the class TestOrcBloomFilters method testMatchesExpandedRange.

@Test
public void testMatchesExpandedRange() {
    Range range = Range.range(BIGINT, 1233L, true, 1235L, true);
    TupleDomainOrcPredicate predicate = TupleDomainOrcPredicate.builder().setBloomFiltersEnabled(true).addColumn(ROOT_COLUMN, Domain.create(ValueSet.ofRanges(range), false)).setDomainCompactionThreshold(100).build();
    ColumnMetadata<ColumnStatistics> matchingStatisticsByColumnIndex = new ColumnMetadata<>(ImmutableList.of(new ColumnStatistics(null, 0, null, new IntegerStatistics(10L, 2000L, null), null, null, null, null, null, null, new Utf8BloomFilterBuilder(1000, 0.01).addLong(1234L).buildBloomFilter())));
    ColumnMetadata<ColumnStatistics> nonMatchingStatisticsByColumnIndex = new ColumnMetadata<>(ImmutableList.of(new ColumnStatistics(null, 0, null, new IntegerStatistics(10L, 2000L, null), null, null, null, null, null, null, new Utf8BloomFilterBuilder(1000, 0.01).addLong(9876L).buildBloomFilter())));
    assertTrue(predicate.matches(1L, matchingStatisticsByColumnIndex));
    assertFalse(predicate.matches(1L, nonMatchingStatisticsByColumnIndex));
}