cboumalh commented on code in PR #52883: URL: https://github.com/apache/spark/pull/52883#discussion_r2627595517
########## sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/tuplesketchesAggregates.scala: ########## @@ -0,0 +1,1895 @@ +/* + * 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.memory.Memory +import org.apache.datasketches.tuple.{Intersection, Sketch, Sketches, Summary, SummaryFactory, SummarySetOperations, Union, UpdatableSketch, UpdatableSketchBuilder, UpdatableSummary} +import org.apache.datasketches.tuple.adouble.{DoubleSummary, DoubleSummaryDeserializer, DoubleSummaryFactory, DoubleSummarySetOperations} +import org.apache.datasketches.tuple.aninteger.{IntegerSummary, IntegerSummaryDeserializer, IntegerSummaryFactory, IntegerSummarySetOperations} +import org.apache.datasketches.tuple.strings.{ArrayOfStringsSummary, ArrayOfStringsSummaryDeserializer, ArrayOfStringsSummaryFactory, ArrayOfStringsSummarySetOperations} + +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.{ExpectsInputTypes, Expression, ExpressionDescription, 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, UnaryLike} +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, and outputs the binary representation of the TupleSketch. + * + * 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( + key: Expression, + summary: Expression, + lgNomEntries: Expression, + mode: Expression, + override val mutableAggBufferOffset: Int, + override val inputAggBufferOffset: Int) + extends TupleSketchAggBase[java.lang.Double, DoubleSummary] + with QuaternaryLike[Expression] + with SummaryAggregateMode { + + // Constructors + def this(key: Expression, summary: Expression) = { + this( + key, + summary, + Literal(ThetaSketchUtils.DEFAULT_LG_NOM_LONGS), + Literal(ThetaSketchUtils.MODE_SUM), + 0, + 0) + } + + def this(key: Expression, summary: Expression, lgNomEntries: Expression) = { + this(key, summary, lgNomEntries, Literal(ThetaSketchUtils.MODE_SUM), 0, 0) + } + + def this(key: Expression, summary: Expression, lgNomEntries: Expression, mode: Expression) = { + this(key, summary, lgNomEntries, mode, 0, 0) + } + + /** + * Override inputTypes to specify key, summary (double/float), lgNomEntries (int), and mode + * (string) parameters. + */ + override def inputTypes: Seq[AbstractDataType] = + Seq( + keyInputTypes, + summaryInputType, + IntegerType, + StringTypeWithCollation(supportsTrimCollation = true)) + + /** + * Override checkInputDataTypes to validate base inputs (key, summary, lgNomEntries) and mode + * parameter. + */ + override def checkInputDataTypes(): TypeCheckResult = { + val defaultCheck = checkBaseInputDataTypes() + if (defaultCheck.isFailure) return defaultCheck + + checkModeParameter() + } + + // Copy constructors required by ImperativeAggregate + override def withNewMutableAggBufferOffset( + newMutableAggBufferOffset: Int): TupleSketchAggDouble = + copy(mutableAggBufferOffset = newMutableAggBufferOffset) + + override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int): TupleSketchAggDouble = + copy(inputAggBufferOffset = newInputAggBufferOffset) + + override protected def withNewChildrenInternal( + newFirst: Expression, + newSecond: Expression, + newThird: Expression, + newFourth: Expression): TupleSketchAggDouble = + copy(key = newFirst, summary = newSecond, lgNomEntries = newThird, mode = newFourth) + + override def first: Expression = key + override def second: Expression = summary + override def third: Expression = lgNomEntries + override def fourth: Expression = mode + + // Overrides for TypedImperativeAggregate + override def prettyName: String = "tuple_sketch_agg_double" + + /** Specifies accepted summary input types (double and float). */ + override protected def summaryInputType: AbstractDataType = + TypeCollection(DoubleType, FloatType) + + /** + * Creates a DoubleSummaryFactory with the configured aggregation mode. + * + * @return + * a DoubleSummaryFactory instance configured with the aggregation mode + */ + override protected def createSummaryFactory(): SummaryFactory[DoubleSummary] = { + val mode = ThetaSketchUtils.getDoubleSummaryMode(modeInput) + new DoubleSummaryFactory(mode) + } + + /** + * Creates DoubleSummarySetOperations for merge operations with the configured mode. + * + * @return + * a DoubleSummarySetOperations instance configured with the aggregation mode + */ + override protected def createSummarySetOperations(): SummarySetOperations[DoubleSummary] = { + val mode = ThetaSketchUtils.getDoubleSummaryMode(modeInput) + new DoubleSummarySetOperations(mode) + } + + /** + * Converts Float inputs to Double, ensuring compatibility with DoubleSummary. + * + * @param input + * the input value to normalize (Float or Double) + * @return + * the normalized Double value + */ + override protected def normalizeSummaryValue(input: Any): java.lang.Double = { + input match { + case d: Double => d + case f: Float => f.toDouble + case _ => + val actualType = input.getClass.getSimpleName + throw QueryExecutionErrors.tupleInvalidSummaryValueType(prettyName, actualType) + } + } + + /** + * Heapify a CompactSketch from the sketch byte array. + * + * @param buffer + * A serialized sketch byte array + * @return + * A CompactSketch instance wrapped with FinalizedTupleSketch + */ + override def deserialize(buffer: Array[Byte]): TupleSketchState[DoubleSummary] = { + if (buffer.nonEmpty) { + FinalizedTupleSketch( + Sketches.heapifySketch(Memory.wrap(buffer), new DoubleSummaryDeserializer())) + } else { + this.createAggregationBuffer() + } + } +} + +/** + * The TupleSketchAggInteger function utilizes a Datasketches TupleSketch instance to count a + * probabilistic approximation of the number of unique values in a given column with associated + * integer type summary values, and outputs the binary representation of the TupleSketch. + * + * 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 (integer 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 TupleSketchAggInteger( + key: Expression, + summary: Expression, + lgNomEntries: Expression, + mode: Expression, + override val mutableAggBufferOffset: Int, + override val inputAggBufferOffset: Int) + extends TupleSketchAggBase[Integer, IntegerSummary] + with QuaternaryLike[Expression] + with SummaryAggregateMode { + + // Constructors + def this(key: Expression, summary: Expression) = { + this( + key, + summary, + Literal(ThetaSketchUtils.DEFAULT_LG_NOM_LONGS), + Literal(ThetaSketchUtils.MODE_SUM), + 0, + 0) + } + + def this(key: Expression, summary: Expression, lgNomEntries: Expression) = { + this(key, summary, lgNomEntries, Literal(ThetaSketchUtils.MODE_SUM), 0, 0) + } + + def this(key: Expression, summary: Expression, lgNomEntries: Expression, mode: Expression) = { + this(key, summary, lgNomEntries, mode, 0, 0) + } + + /** + * Override inputTypes to specify key, summary (integer), lgNomEntries (int), and mode (string) + * parameters. + */ + override def inputTypes: Seq[AbstractDataType] = + Seq( + keyInputTypes, + summaryInputType, + IntegerType, + StringTypeWithCollation(supportsTrimCollation = true)) + + /** + * Override checkInputDataTypes to validate base inputs (key, summary, lgNomEntries) and mode + * parameter. + */ + override def checkInputDataTypes(): TypeCheckResult = { + val defaultCheck = checkBaseInputDataTypes() + if (defaultCheck.isFailure) return defaultCheck + + checkModeParameter() + } + + // Copy constructors required by ImperativeAggregate + override def withNewMutableAggBufferOffset( + newMutableAggBufferOffset: Int): TupleSketchAggInteger = + copy(mutableAggBufferOffset = newMutableAggBufferOffset) + + override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int): TupleSketchAggInteger = + copy(inputAggBufferOffset = newInputAggBufferOffset) + + override protected def withNewChildrenInternal( + newFirst: Expression, + newSecond: Expression, + newThird: Expression, + newFourth: Expression): TupleSketchAggInteger = + copy(key = newFirst, summary = newSecond, lgNomEntries = newThird, mode = newFourth) + + override def first: Expression = key + override def second: Expression = summary + override def third: Expression = lgNomEntries + override def fourth: Expression = mode + + // Overrides for TypedImperativeAggregate + override def prettyName: String = "tuple_sketch_agg_integer" + + /** Specifies accepted summary input types (integer). */ + override protected def summaryInputType: AbstractDataType = + IntegerType + + /** + * Creates an IntegerSummaryFactory with the configured aggregation mode. + * + * @return + * an IntegerSummaryFactory instance configured with the aggregation mode + */ + override protected def createSummaryFactory(): SummaryFactory[IntegerSummary] = { + val mode = ThetaSketchUtils.getIntegerSummaryMode(modeInput) + new IntegerSummaryFactory(mode) + } + + /** + * Creates IntegerSummarySetOperations for merge operations with the configured mode. + * + * @return + * an IntegerSummarySetOperations instance configured with the aggregation mode + */ + override protected def createSummarySetOperations(): SummarySetOperations[IntegerSummary] = { + val mode = ThetaSketchUtils.getIntegerSummaryMode(modeInput) + new IntegerSummarySetOperations(mode, mode) + } + + /** + * Ensures compatibility with IntegerSummary. + * + * @param input + * the input value to normalize (Integer) + * @return + * the normalized Integer value + */ + override protected def normalizeSummaryValue(input: Any): Integer = { + input match { + case i: Int => i + case _ => + val actualType = input.getClass.getSimpleName + throw QueryExecutionErrors.tupleInvalidSummaryValueType(prettyName, actualType) + } + } + + /** + * Heapify a CompactSketch from the sketch byte array. + * + * @param buffer + * A serialized sketch byte array + * @return + * A CompactSketch instance wrapped with FinalizedTupleSketch + */ + override def deserialize(buffer: Array[Byte]): TupleSketchState[IntegerSummary] = { + if (buffer.nonEmpty) { + FinalizedTupleSketch( + Sketches.heapifySketch(Memory.wrap(buffer), new IntegerSummaryDeserializer())) + } else { + this.createAggregationBuffer() + } + } +} + +/** + * The TupleSketchAggString function utilizes a Datasketches TupleSketch instance to count a + * probabilistic approximation of the number of unique values in a given column with associated + * string or string array type summary values, and outputs the binary representation of the + * TupleSketch. + * + * 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 (string or array of strings) to be collected + * @param lgNomEntries + * the log-base-2 of nomEntries decides the number of buckets for the sketch + * @param mutableAggBufferOffset + * offset for mutable aggregation buffer + * @param inputAggBufferOffset + * offset for input aggregation buffer + */ +// scalastyle:off line.size.limit +@ExpressionDescription( + usage = """ + _FUNC_(key, summary, lgNomEntries) - Returns the TupleSketch compact binary representation. + `key` is the expression for unique value counting. + `summary` is the string or array of strings to be collected. + `lgNomEntries` is the log-base-2 of nominal entries, with nominal entries deciding + the number buckets or slots for the TupleSketch. Default is 12. """, + examples = """ + Examples: + > SELECT tuple_sketch_estimate_string(_FUNC_(key, summary, 12)) FROM VALUES (1, 'a'), (1, 'b'), (2, 'c'), (2, 'd'), (3, 'e') tab(key, summary); + 3.0 + """, + group = "agg_funcs", + since = "4.2.0") +// scalastyle:on line.size.limit +case class TupleSketchAggString( + key: Expression, + summary: Expression, + lgNomEntries: Expression, + override val mutableAggBufferOffset: Int, + override val inputAggBufferOffset: Int) + extends TupleSketchAggBase[Array[String], ArrayOfStringsSummary] + with TernaryLike[Expression] { + + // Constructors + def this(key: Expression, summary: Expression) = { + this(key, summary, Literal(ThetaSketchUtils.DEFAULT_LG_NOM_LONGS), 0, 0) + } + + def this(key: Expression, summary: Expression, lgNomEntries: Expression) = { + this(key, summary, lgNomEntries, 0, 0) + } + + /** + * Override inputTypes to specify key, summary (string or array of strings), and lgNomEntries + * (int) parameters. + */ + override def inputTypes: Seq[AbstractDataType] = + Seq(keyInputTypes, summaryInputType, IntegerType) + + /** + * Override checkInputDataTypes to validate base inputs (key, summary, lgNomEntries) only. + */ + override def checkInputDataTypes(): TypeCheckResult = checkBaseInputDataTypes() + + // Copy constructors required by ImperativeAggregate + override def withNewMutableAggBufferOffset( + newMutableAggBufferOffset: Int): TupleSketchAggString = + copy(mutableAggBufferOffset = newMutableAggBufferOffset) + + override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int): TupleSketchAggString = + copy(inputAggBufferOffset = newInputAggBufferOffset) + + override protected def withNewChildrenInternal( + newFirst: Expression, + newSecond: Expression, + newThird: Expression): TupleSketchAggString = + copy(key = newFirst, summary = newSecond, lgNomEntries = newThird) + + override def first: Expression = key + override def second: Expression = summary + override def third: Expression = lgNomEntries + + // Overrides for TypedImperativeAggregate + override def prettyName: String = "tuple_sketch_agg_string" + + /** Specifies accepted summary input types (string or array of strings). */ + override protected def summaryInputType: AbstractDataType = + TypeCollection(StringTypeWithCollation(supportsTrimCollation = true), ArrayType(StringType)) + + /** + * Creates an ArrayOfStringsSummaryFactory. Aggregation mode is not supported here. + * + * @return + * an ArrayOfStringsSummaryFactory instance + */ + override protected def createSummaryFactory(): SummaryFactory[ArrayOfStringsSummary] = { + new ArrayOfStringsSummaryFactory() + } + + /** + * Creates ArrayOfStringsSummarySetOperations for merge operations. Aggregation mode is not + * supported here. + * + * @return + * an ArrayOfStringsSummarySetOperations instance + */ + override protected def createSummarySetOperations() + : SummarySetOperations[ArrayOfStringsSummary] = { + new ArrayOfStringsSummarySetOperations() + } + + /** + * Converts String inputs to String Arrays, ensuring compatibility with ArrayOfStringsSummary. + * + * @param input + * the input value to normalize (UTF8String or ArrayData) + * @return + * the normalized Array[String] value + */ + override protected def normalizeSummaryValue(input: Any): Array[String] = { + input match { + case str: UTF8String => + Array(str.toString) + case arr: ArrayData => + (0 until arr.numElements()) Review Comment: We are skipping the string variant, marking as resolved -- 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]
