cboumalh commented on code in PR #52883: URL: https://github.com/apache/spark/pull/52883#discussion_r2677104174
########## sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/tuplesketchesAggregates.scala: ########## @@ -0,0 +1,1487 @@ +/* + * 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.spark.sql.catalyst.expressions.aggregate + +import org.apache.datasketches.tuple.{Intersection, Sketch, Summary, SummaryFactory, SummarySetOperations, Union, UpdatableSketch, UpdatableSketchBuilder, UpdatableSummary} +import org.apache.datasketches.tuple.adouble.{DoubleSummary, DoubleSummaryFactory, DoubleSummarySetOperations} +import org.apache.datasketches.tuple.aninteger.{IntegerSummary, IntegerSummaryFactory, IntegerSummarySetOperations} + +import org.apache.spark.SparkUnsupportedOperationException +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.analysis.{ExpressionBuilder, TypeCheckResult} +import org.apache.spark.sql.catalyst.analysis.TypeCheckResult.DataTypeMismatch +import org.apache.spark.sql.catalyst.expressions.{Expression, ExpressionDescription, ImplicitCastInputTypes, Literal} +import org.apache.spark.sql.catalyst.expressions.aggregate.TypedImperativeAggregate +import org.apache.spark.sql.catalyst.plans.logical.{FunctionSignature, InputParameter} +import org.apache.spark.sql.catalyst.trees.{BinaryLike, QuaternaryLike, TernaryLike} +import org.apache.spark.sql.catalyst.util.{ArrayData, CollationFactory, ThetaSketchUtils} +import org.apache.spark.sql.errors.QueryExecutionErrors +import org.apache.spark.sql.internal.types.StringTypeWithCollation +import org.apache.spark.sql.types.{AbstractDataType, ArrayType, BinaryType, DataType, DoubleType, FloatType, IntegerType, LongType, StringType, TypeCollection} +import org.apache.spark.unsafe.types.UTF8String + +sealed trait TupleSketchState[S <: Summary] { + def serialize(): Array[Byte] + def eval(): Array[Byte] +} +case class UpdatableTupleSketchBuffer[U, S <: UpdatableSummary[U]](sketch: UpdatableSketch[U, S]) + extends TupleSketchState[S] { + override def serialize(): Array[Byte] = sketch.compact.toByteArray + override def eval(): Array[Byte] = sketch.compact.toByteArray + + /** Returns compact form of the sketch, needed for merge operations that require Sketch type. */ + def compactSketch: Sketch[S] = sketch.compact +} +case class UnionTupleAggregationBuffer[S <: Summary](union: Union[S]) + extends TupleSketchState[S] { + override def serialize(): Array[Byte] = union.getResult.toByteArray + override def eval(): Array[Byte] = union.getResult.toByteArray +} +case class IntersectionTupleAggregationBuffer[S <: Summary](intersection: Intersection[S]) + extends TupleSketchState[S] { + override def serialize(): Array[Byte] = intersection.getResult.toByteArray + override def eval(): Array[Byte] = intersection.getResult.toByteArray +} +case class FinalizedTupleSketch[S <: Summary](sketch: Sketch[S]) extends TupleSketchState[S] { + override def serialize(): Array[Byte] = sketch.toByteArray + override def eval(): Array[Byte] = sketch.toByteArray +} + +/** + * The TupleSketchAggDouble function utilizes a Datasketches TupleSketch instance to count a + * probabilistic approximation of the number of unique values in a given column with associated + * double type summary values that can be aggregated using different modes (sum, min, max, + * alwaysone), and outputs the binary representation of the TupleSketch. + * + * Keys are hashed internally based on their type and value - the same logical value in different + * types (e.g., String("123") and Int(123)) will be treated as distinct keys. However, summary + * value types must be consistent across all calls; mixing types can produce incorrect results or + * precision loss. The value type suffix in the function name (e.g., _double) ensures type safety. + * + * See [[https://datasketches.apache.org/docs/Tuple/TupleSketches.html]] for more information. + * + * @param key + * key expression against which unique counting will occur + * @param summary + * summary expression (double type) against which different mode aggregations will occur + * @param lgNomEntries + * the log-base-2 of nomEntries decides the number of buckets for the sketch + * @param mode + * the aggregation mode for numeric summaries (sum, min, max, alwaysone) + * @param mutableAggBufferOffset + * offset for mutable aggregation buffer + * @param inputAggBufferOffset + * offset for input aggregation buffer + */ +case class TupleSketchAggDouble( Review Comment: @dtenedor Thanks for sharing this suggestion. I'd recommend against this refactoring for the current implementation for these reasons: The base class architecture (TupleSketchAggBase, TupleUnionAggBase, TupleIntersectionAggBase) already does the heavy lifting well as it includes all the sketch logic. The concrete classes are thin wrappers where only ~30 lines per class are actual custom logic: 1. prettyName 2. summaryInputType 3. createSummaryFactory() / createSummarySetOperations() 4. deserialize() with type-specific heapify call The remaining ~100 lines per class are framework-mandated boilerplate (constructors, copy methods, trait implementations) that can't be eliminated even with type classes. I feel like the suggested approach is basically another way of doing the same thing. Please let me know what you think! -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
