Search in sources :

Example 1 with SortedRangeSet

use of com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet in project aws-athena-query-federation by awslabs.

the class JdbcSplitQueryBuilder method toPredicate.

private String toPredicate(String columnName, ValueSet valueSet, ArrowType type, List<TypeAndValue> accumulator) {
    List<String> disjuncts = new ArrayList<>();
    List<Object> singleValues = new ArrayList<>();
    if (valueSet instanceof SortedRangeSet) {
        if (valueSet.isNone() && valueSet.isNullAllowed()) {
            return String.format("(%s IS NULL)", quote(columnName));
        }
        if (valueSet.isNullAllowed()) {
            disjuncts.add(String.format("(%s IS NULL)", quote(columnName)));
        }
        Range rangeSpan = ((SortedRangeSet) valueSet).getSpan();
        if (!valueSet.isNullAllowed() && rangeSpan.getLow().isLowerUnbounded() && rangeSpan.getHigh().isUpperUnbounded()) {
            return String.format("(%s IS NOT NULL)", quote(columnName));
        }
        for (Range range : valueSet.getRanges().getOrderedRanges()) {
            if (range.isSingleValue()) {
                singleValues.add(range.getLow().getValue());
            } else {
                List<String> rangeConjuncts = new ArrayList<>();
                if (!range.getLow().isLowerUnbounded()) {
                    switch(range.getLow().getBound()) {
                        case ABOVE:
                            rangeConjuncts.add(toPredicate(columnName, ">", range.getLow().getValue(), type, accumulator));
                            break;
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(columnName, ">=", range.getLow().getValue(), type, accumulator));
                            break;
                        case BELOW:
                            throw new IllegalArgumentException("Low marker should never use BELOW bound");
                        default:
                            throw new AssertionError("Unhandled bound: " + range.getLow().getBound());
                    }
                }
                if (!range.getHigh().isUpperUnbounded()) {
                    switch(range.getHigh().getBound()) {
                        case ABOVE:
                            throw new IllegalArgumentException("High marker should never use ABOVE bound");
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(columnName, "<=", range.getHigh().getValue(), type, accumulator));
                            break;
                        case BELOW:
                            rangeConjuncts.add(toPredicate(columnName, "<", range.getHigh().getValue(), type, accumulator));
                            break;
                        default:
                            throw new AssertionError("Unhandled bound: " + range.getHigh().getBound());
                    }
                }
                // If rangeConjuncts is null, then the range was ALL, which should already have been checked for
                Preconditions.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(columnName, "=", Iterables.getOnlyElement(singleValues), type, accumulator));
        } else if (singleValues.size() > 1) {
            for (Object value : singleValues) {
                accumulator.add(new TypeAndValue(type, value));
            }
            String values = Joiner.on(",").join(Collections.nCopies(singleValues.size(), "?"));
            disjuncts.add(quote(columnName) + " IN (" + values + ")");
        }
    }
    return "(" + Joiner.on(" OR ").join(disjuncts) + ")";
}
Also used : SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet) ArrayList(java.util.ArrayList) Range(com.amazonaws.athena.connector.lambda.domain.predicate.Range)

Example 2 with SortedRangeSet

use of com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet in project aws-athena-query-federation by awslabs.

the class PropertyGraphHandler method getQueryPartForContraintsMap.

/**
 * Used to generate Gremlin Query part for Constraint Map
 *
 * @param traversal Gremlin Traversal, traversal is updated based on constraints
 *                  map
 * @param hasMap    Constraint Hash Map
 *
 * @return A Gremlin Query Part equivalent to Contraint.
 */
public GraphTraversal getQueryPartForContraintsMap(GraphTraversal traversal, final ReadRecordsRequest recordsRequest) {
    final Map<String, ValueSet> hashMap = recordsRequest.getConstraints().getSummary();
    if (hashMap.size() == 0) {
        return traversal;
    }
    logger.info("readWithContraint: Constaints Map " + hashMap.toString());
    final Set<String> setOfkeys = (Set<String>) (hashMap.keySet());
    for (final String key : setOfkeys) {
        if (hashMap.get(key) instanceof SortedRangeSet) {
            final List<Range> ranges = ((SortedRangeSet) hashMap.get(key)).getOrderedRanges();
            for (final Range range : ranges) {
                if (!range.getLow().isNullValue() && !range.getHigh().isNullValue()) {
                    if (range.getLow().getValue().toString().equals(range.getHigh().getValue().toString())) {
                        traversal = GremlinQueryPreProcessor.generateGremlinQueryPart(traversal, key, range.getLow().getValue().toString(), range.getType(), range.getLow().getBound(), GremlinQueryPreProcessor.Operator.EQUALTO);
                        break;
                    }
                }
                if (!range.getLow().isNullValue()) {
                    traversal = GremlinQueryPreProcessor.generateGremlinQueryPart(traversal, key, range.getLow().getValue().toString(), range.getType(), range.getLow().getBound(), GremlinQueryPreProcessor.Operator.GREATERTHAN);
                }
                if (!range.getHigh().isNullValue()) {
                    traversal = GremlinQueryPreProcessor.generateGremlinQueryPart(traversal, key, range.getHigh().getValue().toString(), range.getType(), range.getHigh().getBound(), GremlinQueryPreProcessor.Operator.LESSTHAN);
                }
            }
        }
        if (hashMap.get(key) instanceof EquatableValueSet) {
            final EquatableValueSet valueSet = ((EquatableValueSet) hashMap.get(key));
            if (valueSet.isWhiteList()) {
                traversal = GremlinQueryPreProcessor.generateGremlinQueryPart(traversal, key, valueSet.getValue(0).toString(), valueSet.getType(), null, GremlinQueryPreProcessor.Operator.EQUALTO);
            } else {
                traversal = GremlinQueryPreProcessor.generateGremlinQueryPart(traversal, key, valueSet.getValue(0).toString(), valueSet.getType(), null, GremlinQueryPreProcessor.Operator.NOTEQUALTO);
            }
        }
    }
    return traversal;
}
Also used : EquatableValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.EquatableValueSet) ValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.ValueSet) SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet) Set(java.util.Set) SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet) EquatableValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.EquatableValueSet) Range(com.amazonaws.athena.connector.lambda.domain.predicate.Range) EquatableValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.EquatableValueSet) ValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.ValueSet)

Example 3 with SortedRangeSet

use of com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet in project aws-athena-query-federation by awslabs.

the class SaphanaQueryStringBuilder method toPredicate.

private String toPredicate(String columnName, ValueSet valueSet, ArrowType type, List<SaphanaQueryStringBuilder.TypeAndValue> accumulator) {
    List<String> disjuncts = new ArrayList<>();
    List<Object> singleValues = new ArrayList<>();
    if (valueSet instanceof SortedRangeSet) {
        if (valueSet.isNone() && valueSet.isNullAllowed()) {
            return String.format("(%s IS NULL)", quote(columnName));
        }
        if (valueSet.isNullAllowed()) {
            disjuncts.add(String.format("(%s IS NULL)", quote(columnName)));
        }
        Range rangeSpan = ((SortedRangeSet) valueSet).getSpan();
        if (!valueSet.isNullAllowed() && rangeSpan.getLow().isLowerUnbounded() && rangeSpan.getHigh().isUpperUnbounded()) {
            return String.format("(%s IS NOT NULL)", quote(columnName));
        }
        for (Range range : valueSet.getRanges().getOrderedRanges()) {
            if (range.isSingleValue()) {
                singleValues.add(range.getLow().getValue());
            } else {
                List<String> rangeConjuncts = new ArrayList<>();
                if (!range.getLow().isLowerUnbounded()) {
                    switch(range.getLow().getBound()) {
                        case ABOVE:
                            rangeConjuncts.add(toPredicate(columnName, ">", range.getLow().getValue(), type, accumulator));
                            break;
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(columnName, ">=", range.getLow().getValue(), type, accumulator));
                            break;
                        case BELOW:
                            throw new IllegalArgumentException("Low marker should never use BELOW bound");
                        default:
                            throw new AssertionError("Unhandled bound: " + range.getLow().getBound());
                    }
                }
                if (!range.getHigh().isUpperUnbounded()) {
                    switch(range.getHigh().getBound()) {
                        case ABOVE:
                            throw new IllegalArgumentException("High marker should never use ABOVE bound");
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(columnName, "<=", range.getHigh().getValue(), type, accumulator));
                            break;
                        case BELOW:
                            rangeConjuncts.add(toPredicate(columnName, "<", range.getHigh().getValue(), type, accumulator));
                            break;
                        default:
                            throw new AssertionError("Unhandled bound: " + range.getHigh().getBound());
                    }
                }
                // If rangeConjuncts is null, then the range was ALL, which should already have been checked for
                Preconditions.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(columnName, "=", Iterables.getOnlyElement(singleValues), type, accumulator));
        } else if (singleValues.size() > 1) {
            for (Object value : singleValues) {
                accumulator.add(new SaphanaQueryStringBuilder.TypeAndValue(type, value));
            }
            String values = Joiner.on(",").join(Collections.nCopies(singleValues.size(), "?"));
            disjuncts.add(quote(columnName) + " IN (" + values + ")");
        }
    }
    return "(" + Joiner.on(" OR ").join(disjuncts) + ")";
}
Also used : SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet) ArrayList(java.util.ArrayList) Range(com.amazonaws.athena.connector.lambda.domain.predicate.Range)

Example 4 with SortedRangeSet

use of com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet in project aws-athena-query-federation by awslabs.

the class CloudwatchRecordHandler method pushDownConstraints.

/**
 * Attempts to push down predicates into Cloudwatch Logs by decorating the Cloudwatch Logs request.
 *
 * @param constraints The constraints for the read as provided by Athena based on the customer's query.
 * @param request The Cloudwatch Logs request to inject predicates to.
 * @return The decorated Cloudwatch Logs request.
 * @note This impl currently only pushing down SortedRangeSet filters (>=, =<, between) on the log time column.
 */
private GetLogEventsRequest pushDownConstraints(Constraints constraints, GetLogEventsRequest request) {
    ValueSet timeConstraint = constraints.getSummary().get(LOG_TIME_FIELD);
    if (timeConstraint instanceof SortedRangeSet && !timeConstraint.isNullAllowed()) {
        // SortedRangeSet is how >, <, between is represented which are easiest and most common when
        // searching logs so we attempt to push that down here as an optimization. SQL can represent complex
        // overlapping ranges which Cloudwatch can not support so this is not a replacement for applying
        // constraints using the ConstraintEvaluator.
        Range basicPredicate = ((SortedRangeSet) timeConstraint).getSpan();
        if (!basicPredicate.getLow().isNullValue()) {
            Long lowerBound = (Long) basicPredicate.getLow().getValue();
            request.setStartTime(lowerBound);
        }
        if (!basicPredicate.getHigh().isNullValue()) {
            Long upperBound = (Long) basicPredicate.getHigh().getValue();
            request.setEndTime(upperBound);
        }
    }
    return request;
}
Also used : SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet) AtomicLong(java.util.concurrent.atomic.AtomicLong) Range(com.amazonaws.athena.connector.lambda.domain.predicate.Range) ValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.ValueSet)

Example 5 with SortedRangeSet

use of com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet in project aws-athena-query-federation by awslabs.

the class DDBPredicateUtils method generateSingleColumnFilter.

/**
 * Generates a filter expression for a single column given a {@link ValueSet} predicate for that column.
 *
 * @param originalColumnName the column name
 * @param predicate the associated predicate
 * @param accumulator the value accumulator to add values to
 * @param valueNameProducer the value name producer to generate value aliases with
 * @param recordMetadata object containing any necessary metadata from the glue table
 * @return the generated filter expression
 */
public static String generateSingleColumnFilter(String originalColumnName, ValueSet predicate, List<AttributeValue> accumulator, IncrementingValueNameProducer valueNameProducer, DDBRecordMetadata recordMetadata) {
    String columnName = aliasColumn(originalColumnName);
    if (predicate.isNone()) {
        return "(attribute_not_exists(" + columnName + ") OR " + toPredicate(originalColumnName, "=", null, accumulator, valueNameProducer.getNext(), recordMetadata) + ")";
    }
    if (predicate.isAll()) {
        return "(attribute_exists(" + columnName + ") AND " + toPredicate(originalColumnName, "<>", null, accumulator, valueNameProducer.getNext(), recordMetadata) + ")";
    }
    List<String> disjuncts = new ArrayList<>();
    List<Object> singleValues = new ArrayList<>();
    boolean isWhitelist = true;
    if (predicate instanceof SortedRangeSet) {
        for (Range range : predicate.getRanges().getOrderedRanges()) {
            // Already checked
            checkState(!range.isAll());
            if (range.isSingleValue()) {
                singleValues.add(range.getLow().getValue());
            } else {
                List<String> rangeConjuncts = new ArrayList<>();
                if (!range.getLow().isLowerUnbounded()) {
                    switch(range.getLow().getBound()) {
                        case ABOVE:
                            rangeConjuncts.add(toPredicate(originalColumnName, ">", range.getLow().getValue(), accumulator, valueNameProducer.getNext(), recordMetadata));
                            break;
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(originalColumnName, ">=", range.getLow().getValue(), accumulator, valueNameProducer.getNext(), recordMetadata));
                            break;
                        case BELOW:
                            throw new IllegalArgumentException("Low marker should never use BELOW bound");
                        default:
                            throw new AssertionError("Unhandled lower bound: " + range.getLow().getBound());
                    }
                }
                if (!range.getHigh().isUpperUnbounded()) {
                    switch(range.getHigh().getBound()) {
                        case ABOVE:
                            throw new IllegalArgumentException("High marker should never use ABOVE bound");
                        case EXACTLY:
                            rangeConjuncts.add(toPredicate(originalColumnName, "<=", range.getHigh().getValue(), accumulator, valueNameProducer.getNext(), recordMetadata));
                            break;
                        case BELOW:
                            rangeConjuncts.add(toPredicate(originalColumnName, "<", range.getHigh().getValue(), accumulator, valueNameProducer.getNext(), recordMetadata));
                            break;
                        default:
                            throw new AssertionError("Unhandled upper bound: " + range.getHigh().getBound());
                    }
                }
                // If rangeConjuncts is null, then the range was ALL, which should already have been checked for
                checkState(!rangeConjuncts.isEmpty());
                disjuncts.add("(" + AND_JOINER.join(rangeConjuncts) + ")");
            }
        }
    } else {
        EquatableValueSet equatablePredicate = (EquatableValueSet) predicate;
        isWhitelist = equatablePredicate.isWhiteList();
        long valueCount = equatablePredicate.getValueBlock().getRowCount();
        for (int i = 0; i < valueCount; i++) {
            singleValues.add(equatablePredicate.getValue(i));
        }
    }
    // Add back all of the possible single values either as an equality or an IN predicate
    if (singleValues.size() == 1) {
        disjuncts.add(toPredicate(originalColumnName, isWhitelist ? "=" : "<>", getOnlyElement(singleValues), accumulator, valueNameProducer.getNext(), recordMetadata));
    } else if (singleValues.size() > 1) {
        for (Object value : singleValues) {
            bindValue(originalColumnName, value, accumulator, recordMetadata);
        }
        String values = COMMA_JOINER.join(Stream.generate(valueNameProducer::getNext).limit(singleValues.size()).collect(toImmutableList()));
        disjuncts.add((isWhitelist ? "" : "NOT ") + columnName + " IN (" + values + ")");
    }
    // at this point we should have some disjuncts
    checkState(!disjuncts.isEmpty());
    // add nullability disjuncts
    if (predicate.isNullAllowed()) {
        disjuncts.add("attribute_not_exists(" + columnName + ") OR " + toPredicate(originalColumnName, "=", null, accumulator, valueNameProducer.getNext(), recordMetadata));
    }
    // DDB doesn't like redundant parentheses
    if (disjuncts.size() == 1) {
        return disjuncts.get(0);
    }
    return "(" + OR_JOINER.join(disjuncts) + ")";
}
Also used : EquatableValueSet(com.amazonaws.athena.connector.lambda.domain.predicate.EquatableValueSet) ArrayList(java.util.ArrayList) Range(com.amazonaws.athena.connector.lambda.domain.predicate.Range) SortedRangeSet(com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet)

Aggregations

Range (com.amazonaws.athena.connector.lambda.domain.predicate.Range)9 SortedRangeSet (com.amazonaws.athena.connector.lambda.domain.predicate.SortedRangeSet)9 ArrayList (java.util.ArrayList)7 EquatableValueSet (com.amazonaws.athena.connector.lambda.domain.predicate.EquatableValueSet)3 ValueSet (com.amazonaws.athena.connector.lambda.domain.predicate.ValueSet)3 Split (com.amazonaws.athena.connector.lambda.domain.Split)1 GetMetricDataRequest (com.amazonaws.services.cloudwatch.model.GetMetricDataRequest)1 Metric (com.amazonaws.services.cloudwatch.model.Metric)1 MetricDataQuery (com.amazonaws.services.cloudwatch.model.MetricDataQuery)1 MetricStat (com.amazonaws.services.cloudwatch.model.MetricStat)1 Date (java.util.Date)1 List (java.util.List)1 Set (java.util.Set)1 AtomicLong (java.util.concurrent.atomic.AtomicLong)1