spark AggregateInPandasExec 源码

spark AggregateInPandasExec 代码

文件路径：/sql/core/src/main/scala/org/apache/spark/sql/execution/python/AggregateInPandasExec.scala

/*
 * 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.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.{AllTuples, ClusteredDistribution, Distribution, Partitioning}
import org.apache.spark.sql.execution.{GroupedIterator, SparkPlan, UnaryExecNode}
import org.apache.spark.sql.execution.aggregate.UpdatingSessionsIterator
import org.apache.spark.sql.types.{DataType, StructField, StructType}
import org.apache.spark.sql.util.ArrowUtils
import org.apache.spark.util.Utils

/**
 * Physical node for aggregation with group aggregate Pandas UDF.
 *
 * This plan works by sending the necessary (projected) input grouped data as Arrow record batches
 * to the python worker, the python worker invokes the UDF and sends the results to the executor,
 * finally the executor evaluates any post-aggregation expressions and join the result with the
 * grouped key.
 */
case class AggregateInPandasExec(
    groupingExpressions: Seq[NamedExpression],
    udfExpressions: Seq[PythonUDF],
    resultExpressions: Seq[NamedExpression],
    child: SparkPlan)
  extends UnaryExecNode {

  override val output: Seq[Attribute] = resultExpressions.map(_.toAttribute)

  override def outputPartitioning: Partitioning = child.outputPartitioning

  override def producedAttributes: AttributeSet = AttributeSet(output)

  val sessionWindowOption = groupingExpressions.find { p =>
    p.metadata.contains(SessionWindow.marker)
  }

  val groupingWithoutSessionExpressions = sessionWindowOption match {
    case Some(sessionExpression) =>
      groupingExpressions.filterNot { p => p.semanticEquals(sessionExpression) }

    case None => groupingExpressions
  }

  override def requiredChildDistribution: Seq[Distribution] = {
    if (groupingExpressions.isEmpty) {
      AllTuples :: Nil
    } else {
      ClusteredDistribution(groupingExpressions) :: Nil
    }
  }

  override def requiredChildOrdering: Seq[Seq[SortOrder]] = sessionWindowOption match {
    case Some(sessionExpression) =>
      Seq((groupingWithoutSessionExpressions ++ Seq(sessionExpression))
        .map(SortOrder(_, Ascending)))

    case None => Seq(groupingExpressions.map(SortOrder(_, Ascending)))
  }

  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(!_.exists(_.isInstanceOf[PythonUDF])))
        (ChainedPythonFunctions(Seq(udf.func)), udf.children)
    }
  }

  override protected def doExecute(): RDD[InternalRow] = {
    val inputRDD = child.execute()

    val sessionLocalTimeZone = conf.sessionLocalTimeZone
    val pythonRunnerConf = ArrowUtils.getPythonRunnerConfMap(conf)

    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))) {
          allInputs.indexWhere(_.semanticEquals(e))
        } else {
          allInputs += e
          dataTypes += e.dataType
          allInputs.length - 1
        }
      }.toArray
    }.toArray

    // Schema of input rows to the python runner
    val aggInputSchema = StructType(dataTypes.zipWithIndex.map { case (dt, i) =>
      StructField(s"_$i", dt)
    }.toSeq)

    // Map grouped rows to ArrowPythonRunner results, Only execute if partition is not empty
    inputRDD.mapPartitionsInternal { iter => if (iter.isEmpty) iter else {
      // If we have session window expression in aggregation, we wrap iterator with
      // UpdatingSessionIterator to calculate sessions for input rows and update
      // rows' session column, so that further aggregations can aggregate input rows
      // for the same session.
      val newIter: Iterator[InternalRow] = mayAppendUpdatingSessionIterator(iter)
      val prunedProj = UnsafeProjection.create(allInputs.toSeq, child.output)

      val groupedItr = if (groupingExpressions.isEmpty) {
        // Use an empty unsafe row as a place holder for the grouping key
        Iterator((new UnsafeRow(), newIter))
      } else {
        GroupedIterator(newIter, groupingExpressions, child.output)
      }
      val grouped = groupedItr.map { case (key, rows) =>
        (key, rows.map(prunedProj))
      }

      val context = TaskContext.get()

      // The queue used to buffer input rows so we can drain it to
      // combine input with output from Python.
      val queue = HybridRowQueue(context.taskMemoryManager(),
        new File(Utils.getLocalDir(SparkEnv.get.conf)), groupingExpressions.length)
      context.addTaskCompletionListener[Unit] { _ =>
        queue.close()
      }

      // Add rows to queue to join later with the result.
      val projectedRowIter = grouped.map { case (groupingKey, rows) =>
        queue.add(groupingKey.asInstanceOf[UnsafeRow])
        rows
      }

      val columnarBatchIter = new ArrowPythonRunner(
        pyFuncs,
        PythonEvalType.SQL_GROUPED_AGG_PANDAS_UDF,
        argOffsets,
        aggInputSchema,
        sessionLocalTimeZone,
        pythonRunnerConf).compute(projectedRowIter, context.partitionId(), context)

      val joinedAttributes =
        groupingExpressions.map(_.toAttribute) ++ udfExpressions.map(_.resultAttribute)
      val joined = new JoinedRow
      val resultProj = UnsafeProjection.create(resultExpressions, joinedAttributes)

      columnarBatchIter.map(_.rowIterator.next()).map { aggOutputRow =>
        val leftRow = queue.remove()
        val joinedRow = joined(leftRow, aggOutputRow)
        resultProj(joinedRow)
      }
    }}
  }

  override protected def withNewChildInternal(newChild: SparkPlan): SparkPlan =
    copy(child = newChild)


  private def mayAppendUpdatingSessionIterator(
      iter: Iterator[InternalRow]): Iterator[InternalRow] = {
    val newIter = sessionWindowOption match {
      case Some(sessionExpression) =>
        val inMemoryThreshold = conf.windowExecBufferInMemoryThreshold
        val spillThreshold = conf.windowExecBufferSpillThreshold

        new UpdatingSessionsIterator(iter, groupingWithoutSessionExpressions, sessionExpression,
          child.output, inMemoryThreshold, spillThreshold)

      case None => iter
    }

    newIter
  }
}

相关信息

spark 源码目录

相关文章

spark ApplyInPandasWithStatePythonRunner 源码

spark ApplyInPandasWithStateWriter 源码

spark ArrowEvalPythonExec 源码

spark ArrowPythonRunner 源码

spark AttachDistributedSequenceExec 源码

spark BatchEvalPythonExec 源码

spark CoGroupedArrowPythonRunner 源码

spark EvalPythonExec 源码

spark EvaluatePython 源码

spark ExtractPythonUDFs 源码