Github user hvanhovell commented on a diff in the pull request:
https://github.com/apache/spark/pull/14136#discussion_r89652990
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/Percentile.scala
---
@@ -0,0 +1,292 @@
+/*
+ * 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.spark.sql.AnalysisException
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.analysis.TypeCheckResult
+import org.apache.spark.sql.catalyst.expressions._
+import
org.apache.spark.sql.catalyst.expressions.aggregate.Percentile.Countings
+import org.apache.spark.sql.catalyst.util._
+import org.apache.spark.sql.types._
+import org.apache.spark.unsafe.Platform.BYTE_ARRAY_OFFSET
+import org.apache.spark.util.collection.OpenHashMap
+
+
+/**
+ * The Percentile aggregate function returns the exact percentile(s) of
numeric column `expr` at
+ * the given percentage(s) with value range in [0.0, 1.0].
+ *
+ * The operator is bound to the slower sort based aggregation path because
the number of elements
+ * and their partial order cannot be determined in advance. Therefore we
have to store all the
+ * elements in memory, and that too many elements can cause GC paused and
eventually OutOfMemory
+ * Errors.
+ *
+ * @param child child expression that produce numeric column value with
`child.eval(inputRow)`
+ * @param percentageExpression Expression that represents a single
percentage value or an array of
+ * percentage values. Each percentage value
must be in the range
+ * [0.0, 1.0].
+ */
+@ExpressionDescription(
+ usage =
+ """
+ _FUNC_(col, percentage) - Returns the exact percentile value of
numeric column `col` at the
+ given percentage. The value of percentage must be between 0.0 and
1.0.
+
+ _FUNC_(col, array(percentage1 [, percentage2]...)) - Returns the
exact percentile value array
+ of numeric column `col` at the given percentage(s). Each value of
the percentage array must
+ be between 0.0 and 1.0.
+ """)
+case class Percentile(
+ child: Expression,
+ percentageExpression: Expression,
+ mutableAggBufferOffset: Int = 0,
+ inputAggBufferOffset: Int = 0) extends
TypedImperativeAggregate[Countings] {
+
+ def this(child: Expression, percentageExpression: Expression) = {
+ this(child, percentageExpression, 0, 0)
+ }
+
+ override def prettyName: String = "percentile"
+
+ override def withNewMutableAggBufferOffset(newMutableAggBufferOffset:
Int): Percentile =
+ copy(mutableAggBufferOffset = newMutableAggBufferOffset)
+
+ override def withNewInputAggBufferOffset(newInputAggBufferOffset: Int):
Percentile =
+ copy(inputAggBufferOffset = newInputAggBufferOffset)
+
+ // Mark as lazy so that percentageExpression is not evaluated during
tree transformation.
+ private lazy val (returnPercentileArray: Boolean, percentages:
Seq[Number]) =
+ evalPercentages(percentageExpression)
+
+ override def children: Seq[Expression] = child :: percentageExpression
:: Nil
+
+ // Returns null for empty inputs
+ override def nullable: Boolean = true
+
+ override def dataType: DataType =
+ if (returnPercentileArray) ArrayType(DoubleType) else DoubleType
+
+ override def inputTypes: Seq[AbstractDataType] =
+ Seq(NumericType, TypeCollection(NumericType, ArrayType))
+
+ override def checkInputDataTypes(): TypeCheckResult =
+ TypeUtils.checkForNumericExpr(child.dataType, "function percentile")
+
+ override def createAggregationBuffer(): Countings = {
+ // Initialize new Countings instance here.
+ Countings()
+ }
+
+ private def evalPercentages(expr: Expression): (Boolean, Seq[Number]) = {
+ val (isArrayType, values) = (expr.dataType, expr.eval()) match {
+ case (_, n: Number) => (false, Array(n))
+ case (_, d: Decimal) => (false,
Array(d.toDouble.asInstanceOf[Number]))
+ case (ArrayType(baseType: NumericType, _), arrayData: ArrayData) =>
+ val numericArray = arrayData.toObjectArray(baseType)
+ (true, numericArray.map { x =>
+
baseType.numeric.toDouble(x.asInstanceOf[baseType.InternalType]).asInstanceOf[Number]
+ })
+ case other =>
+ throw new AnalysisException(s"Invalid data type ${other._1} for
parameter percentage")
+ }
+
+ require(values.forall(value => value.doubleValue() >= 0.0 &&
value.doubleValue() <= 1.0),
+ s"Percentage values must be between 0.0 and 1.0, current values =
${values.mkString(", ")}")
+
+ (isArrayType, values)
+ }
+
+ override def update(buffer: Countings, input: InternalRow): Unit = {
+ val key = child.eval(input).asInstanceOf[Number]
+ buffer.add(key)
+ }
+
+ override def merge(buffer: Countings, other: Countings): Unit = {
+ buffer.merge(other)
+ }
+
+ override def eval(buffer: Countings): Any = {
+ generateOutput(buffer.getPercentiles(percentages))
+ }
+
+ private def generateOutput(results: Seq[Double]): Any = {
+ if (results.isEmpty) {
+ null
+ } else if (returnPercentileArray) {
+ new GenericArrayData(results)
+ } else {
+ results.head
+ }
+ }
+
+ override def serialize(obj: Countings): Array[Byte] = {
+ Percentile.serializer.serialize(obj, child.dataType)
+ }
+
+ override def deserialize(bytes: Array[Byte]): Countings = {
+ Percentile.serializer.deserialize(bytes, child.dataType)
+ }
+}
+
+object Percentile {
+ object Countings {
+ def apply(): Countings = Countings(new OpenHashMap[Number, Long])
+
+ def apply(counts: OpenHashMap[Number, Long]): Countings = new
Countings(counts)
+ }
+
+ /**
+ * A class that stores the numbers and their counts, used to support
[[Percentile]] function.
+ */
+ class Countings(val counts: OpenHashMap[Number, Long]) extends
Serializable {
+ /**
+ * Insert a key into countings map.
+ */
+ def add(key: Number): Unit = {
+ // Null values are ignored in countings.
+ if (key != null) {
+ counts.changeValue(key, 1L, _ + 1L)
+ }
+ }
+
+ /**
+ * In place merges in another Countings.
+ */
+ def merge(other: Countings): Unit = {
+ other.counts.foreach { pair =>
+ counts.changeValue(pair._1, pair._2, _ + pair._2)
+ }
+ }
+
+ /**
+ * Get the percentile value for every percentile in `percentages`.
+ */
+ def getPercentiles(percentages: Seq[Number]): Seq[Double] = {
+ if (counts.isEmpty) {
+ return Seq.empty
+ }
+
+ val sortedCounts = counts.toSeq.sortBy(_._1)(new Ordering[Number]() {
+ override def compare(a: Number, b: Number): Int =
+ scala.math.signum(a.doubleValue() - b.doubleValue()).toInt
+ })
+ val aggreCounts = sortedCounts.scanLeft(sortedCounts.head._1, 0L) {
+ (k1: (Number, Long), k2: (Number, Long)) => (k2._1, k1._2 + k2._2)
+ }.tail
+ val maxPosition = aggreCounts.last._2 - 1
+
+ percentages.map { percentile =>
+ getPercentile(aggreCounts, maxPosition *
percentile.doubleValue()).doubleValue()
+ }
+ }
+
+ /**
+ * Get the percentile value.
+ */
+ private def getPercentile(aggreCounts: Seq[(Number, Long)], position:
Double): Number = {
+ // We may need to do linear interpolation to get the exact percentile
+ val lower = position.floor
+ val higher = position.ceil
+
+ // Linear search since this won't take much time from the total
execution anyway
--- End diff --
This was taken from Hive `UDAFPercentile`. It is fine if you do that, but
please acknowledge that you have done so by adding a line of documentation. See
this for example:
https://github.com/apache/spark/blob/master/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/windowExpressions.scala#L524
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
