Github user HyukjinKwon commented on a diff in the pull request: https://github.com/apache/spark/pull/21082#discussion_r184875120 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/python/WindowInPandasExec.scala --- @@ -0,0 +1,174 @@ +/* + * 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.execution.python + +import java.io.File + +import scala.collection.JavaConverters._ +import scala.collection.mutable.ArrayBuffer + +import org.apache.spark.{SparkEnv, TaskContext} +import org.apache.spark.api.python.{ChainedPythonFunctions, PythonEvalType} +import org.apache.spark.rdd.RDD +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions._ +import org.apache.spark.sql.catalyst.plans.physical._ +import org.apache.spark.sql.execution.{GroupedIterator, SparkPlan, UnaryExecNode} +import org.apache.spark.sql.types.{DataType, StructField, StructType} +import org.apache.spark.util.Utils + +case class WindowInPandasExec( + windowExpression: Seq[NamedExpression], + partitionSpec: Seq[Expression], + orderSpec: Seq[SortOrder], + child: SparkPlan) extends UnaryExecNode { + + override def output: Seq[Attribute] = + child.output ++ windowExpression.map(_.toAttribute) + + override def requiredChildDistribution: Seq[Distribution] = { + if (partitionSpec.isEmpty) { + // Only show warning when the number of bytes is larger than 100 MB? + logWarning("No Partition Defined for Window operation! Moving all data to a single " + + "partition, this can cause serious performance degradation.") + AllTuples :: Nil + } else ClusteredDistribution(partitionSpec) :: Nil + } + + override def requiredChildOrdering: Seq[Seq[SortOrder]] = + Seq(partitionSpec.map(SortOrder(_, Ascending)) ++ orderSpec) + + override def outputOrdering: Seq[SortOrder] = child.outputOrdering + + override def outputPartitioning: Partitioning = child.outputPartitioning + + private def collectFunctions(udf: PythonUDF): (ChainedPythonFunctions, Seq[Expression]) = { + udf.children match { + case Seq(u: PythonUDF) => + val (chained, children) = collectFunctions(u) + (ChainedPythonFunctions(chained.funcs ++ Seq(udf.func)), children) + case children => + // There should not be any other UDFs, or the children can't be evaluated directly. + assert(children.forall(_.find(_.isInstanceOf[PythonUDF]).isEmpty)) + (ChainedPythonFunctions(Seq(udf.func)), udf.children) + } + } + + /** + * Create the resulting projection. + * + * This method uses Code Generation. It can only be used on the executor side. + * + * @param expressions unbound ordered function expressions. + * @return the final resulting projection. + */ + private[this] def createResultProjection(expressions: Seq[Expression]): UnsafeProjection = { + val references = expressions.zipWithIndex.map{ case (e, i) => + // Results of window expressions will be on the right side of child's output + BoundReference(child.output.size + i, e.dataType, e.nullable) + } + val unboundToRefMap = expressions.zip(references).toMap + val patchedWindowExpression = windowExpression.map(_.transform(unboundToRefMap)) + UnsafeProjection.create( + child.output ++ patchedWindowExpression, + child.output) + } + + protected override def doExecute(): RDD[InternalRow] = { + val inputRDD = child.execute() + + val bufferSize = inputRDD.conf.getInt("spark.buffer.size", 65536) + val reuseWorker = inputRDD.conf.getBoolean("spark.python.worker.reuse", defaultValue = true) + val sessionLocalTimeZone = conf.sessionLocalTimeZone + val pandasRespectSessionTimeZone = conf.pandasRespectSessionTimeZone + + // Extract window expressions and window functions + val expressions = windowExpression.flatMap { e => + e.collect { + case e: WindowExpression => e + } + } + val udfExpressions = expressions.map(_.windowFunction.asInstanceOf[PythonUDF]) + + val (pyFuncs, inputs) = udfExpressions.map(collectFunctions).unzip + + // Filter child output attributes down to only those that are UDF inputs. + // Also eliminate duplicate UDF inputs. + val allInputs = new ArrayBuffer[Expression] + val dataTypes = new ArrayBuffer[DataType] + val argOffsets = inputs.map { input => + input.map { e => + if (allInputs.exists(_.semanticEquals(e))) { --- End diff -- don't feel strongly but I would have used `case` with if.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org