steFaiz commented on code in PR #7828: URL: https://github.com/apache/paimon/pull/7828#discussion_r3239413679
########## paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/source/aggregate/AggregatePushDownUtils.java: ########## @@ -0,0 +1,304 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.paimon.flink.source.aggregate; + +import org.apache.paimon.CoreOptions; +import org.apache.paimon.flink.LogicalTypeConversion; +import org.apache.paimon.partition.PartitionPredicate; +import org.apache.paimon.predicate.Predicate; +import org.apache.paimon.table.FileStoreTable; +import org.apache.paimon.table.source.DataSplit; +import org.apache.paimon.table.source.ReadBuilder; +import org.apache.paimon.table.source.Split; +import org.apache.paimon.types.DataField; +import org.apache.paimon.types.DataType; +import org.apache.paimon.types.RowType; +import org.apache.paimon.utils.Projection; + +import org.apache.flink.table.expressions.AggregateExpression; + +import javax.annotation.Nullable; + +import java.util.ArrayList; +import java.util.List; +import java.util.Optional; +import java.util.OptionalLong; +import java.util.Set; +import java.util.stream.Collectors; + +import static org.apache.paimon.table.source.PushDownUtils.minmaxAvailable; + +/** + * Utilities for aggregate push down. + * + * <p>This class references Spark's implementation: {@code + * org.apache.paimon.spark.aggregate.AggregatePushDownUtils} and related classes. + */ +public class AggregatePushDownUtils { + + public static Optional<PushedAggregateResult> tryPushdownAggregation( + FileStoreTable table, + @Nullable Predicate predicate, + @Nullable PartitionPredicate partitionPredicate, + @Nullable int[][] projectFields, + List<int[]> groupingSets, + List<AggregateExpression> aggregateExpressions, + org.apache.flink.table.types.DataType producedDataType) { + if (groupingSets.size() != 1) { + return Optional.empty(); + } + + // reject nested projection + int[] fieldIndexMapping = createFieldIndexMapping(table, projectFields); + if (fieldIndexMapping == null) { + return Optional.empty(); + } + + // groupingSet contains the index within the projected fields, + // so we have to translate grouping fields index to original field index + int[] originalGrouping = translateFieldIndexes(groupingSets.get(0), fieldIndexMapping); + if (originalGrouping == null) { + return Optional.empty(); + } + + if (originalGrouping.length > 0 + && !groupingFieldsArePartitionFields(table, originalGrouping)) { + return Optional.empty(); + } + + List<LocalAggregator.Aggregate> aggregates = + extractAggregates(table, fieldIndexMapping, aggregateExpressions); + if (aggregates == null) { + return Optional.empty(); + } + + List<DataSplit> dataSplits = planSplits(table, predicate, partitionPredicate, aggregates); + if (dataSplits == null) { + return Optional.empty(); + } + + LocalAggregator aggregator = new LocalAggregator(table, originalGrouping, aggregates); + for (DataSplit dataSplit : dataSplits) { + aggregator.update(dataSplit); + } + + // we should check the result row type equals to producedDataType + RowType producedRowType = toPaimonRowType(producedDataType); + if (producedRowType == null || !isCompatible(aggregator.resultRowType(), producedRowType)) { + return Optional.empty(); + } + return Optional.of(new PushedAggregateResult(aggregator.result(), producedRowType)); + } + + @Nullable + private static RowType toPaimonRowType(org.apache.flink.table.types.DataType producedDataType) { + if (!(producedDataType.getLogicalType() + instanceof org.apache.flink.table.types.logical.RowType)) { + return null; + } + return LogicalTypeConversion.toDataType( + (org.apache.flink.table.types.logical.RowType) producedDataType.getLogicalType()); + } + + private static boolean isCompatible(RowType actualRowType, RowType producedRowType) { + if (actualRowType.getFieldCount() != producedRowType.getFieldCount()) { + return false; + } + + for (int i = 0; i < actualRowType.getFieldCount(); i++) { + DataType actualType = actualRowType.getTypeAt(i); + DataType producedType = producedRowType.getTypeAt(i); + if (!actualType.equalsIgnoreNullable(producedType) + || (actualType.isNullable() && !producedType.isNullable())) { + return false; + } + } + return true; + } + + @Nullable + private static List<DataSplit> planSplits( + FileStoreTable table, + @Nullable Predicate predicate, + @Nullable PartitionPredicate partitionPredicate, + List<LocalAggregator.Aggregate> aggregates) { + Set<String> minMaxColumns = + aggregates.stream() + .filter(LocalAggregator.Aggregate::requiresMinMaxStats) + .map(LocalAggregator.Aggregate::fieldName) + .collect(Collectors.toSet()); + if (!minMaxColumns.isEmpty() + && (CoreOptions.fromMap(table.options()).deletionVectorsEnabled() + || !table.primaryKeys().isEmpty())) { + return null; + } + + ReadBuilder readBuilder = + table.newReadBuilder() + .withFilter(predicate) + .withPartitionFilter(partitionPredicate); + if (minMaxColumns.isEmpty()) { + readBuilder.dropStats(); + } + + List<Split> splits = readBuilder.newScan().plan().splits(); + List<DataSplit> dataSplits = new ArrayList<>(splits.size()); + for (Split split : splits) { + if (!(split instanceof DataSplit)) { + return null; + } + + OptionalLong mergedRowCount = split.mergedRowCount(); + if (!mergedRowCount.isPresent()) { + return null; + } + + if (!minMaxColumns.isEmpty() && !minmaxAvailable(split, minMaxColumns)) { + return null; + } + + dataSplits.add((DataSplit) split); + } + return dataSplits; + } + + @Nullable + private static List<LocalAggregator.Aggregate> extractAggregates( + FileStoreTable table, + int[] fieldIndexMapping, + List<AggregateExpression> aggregateExpressions) { + List<LocalAggregator.Aggregate> aggregates = new ArrayList<>(aggregateExpressions.size()); + for (AggregateExpression aggregateExpression : aggregateExpressions) { + if (aggregateExpression.isDistinct() + || aggregateExpression.isApproximate() + || aggregateExpression.getFilterExpression().isPresent()) { + return null; + } + + String functionName = aggregateExpression.getFunctionDefinition().getClass().getName(); + if (isCountStar(functionName)) { + aggregates.add(LocalAggregator.Aggregate.count()); + } else if (isMin(functionName) || isMax(functionName)) { + if (aggregateExpression.getArgs().size() != 1) { + return null; + } + + int originalFieldIndex = + toOriginalFieldIndex( + fieldIndexMapping, + aggregateExpression.getArgs().get(0).getFieldIndex()); Review Comment: Thanks for your remind! During implementation, I referred to Flink's original agg pushdown interface, which only supports pushing down simple fields: <img width="1224" height="334" alt="image" src="https://github.com/user-attachments/assets/64c5b393-1815-4e4f-9241-b550f406c7c0"; /> For example, if the sql is: ```sql SELECT MIN(f0 * 2) FROM paimon ``` The plan will be: ```text LocalHashAggregate(Partial_MIN($f0)) +- Calc(select=[(f0 * 2) AS $f0]) +- TableSourceScan(project=[f0]) ``` Since Flink do not consider these situations, the args of `AggregateExpression` is hard-coded as `FieldReferenceExpression`: <img width="1102" height="276" alt="image" src="https://github.com/user-attachments/assets/21c9e59e-a12f-4718-8070-fccb0d8efb16"; /> I've added a test for these cases! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
