Question

以下代码引发异常，原因是：java.lang.UnsupportedOperationException：没有模式的行上的fieldIndex未定义。当在使用ExpressionEncoder，groupedByKey和flatMap调用数据框上的groupByKey和flatMap之后返回的数据框上出现时，就会发生这种情况。

逻辑流程： originalDf-> groupByKey-> flatMap-> groupByKey-> flatMap->显示

   import org.apache.spark.sql.catalyst.encoders.RowEncoder
import org.apache.spark.sql.{Row, SparkSession}
import org.apache.spark.sql.types.{ IntegerType, StructField, StructType}

import scala.collection.mutable.ListBuffer



  object Test {

    def main(args: Array[String]): Unit = {

      val values = List(List("1", "One") ,List("1", "Two") ,List("2", "Three"),List("2","4")).map(x =>(x(0), x(1)))
      val session = SparkSession.builder.config("spark.master", "local").getOrCreate
      import session.implicits._
      val dataFrame = values.toDF


      dataFrame.show()
      dataFrame.printSchema()

      val newSchema = StructType(dataFrame.schema.fields
        ++ Array(
        StructField("Count", IntegerType, false)
      )
      )

      val expr = RowEncoder.apply(newSchema)

      val tranform =  dataFrame.groupByKey(row => row.getAs[String]("_1")).flatMapGroups((key, inputItr) => {
        val inputSeq = inputItr.toSeq

        val length = inputSeq.size
        var listBuff = new ListBuffer[Row]()
        var counter : Int= 0
        for(i <- 0 until(length))
        {
          counter+=1

        }

        for(i <- 0 until length ) {
          var x = inputSeq(i)
          listBuff += Row.fromSeq(x.toSeq ++ Array[Int](counter))
        }
        listBuff.iterator
      })(expr)

      tranform.show

      val newSchema1 = StructType(tranform.schema.fields
        ++ Array(
        StructField("Count1", IntegerType, false)
      )
      )
      val expr1 = RowEncoder.apply(newSchema1)
      val tranform2 =  tranform.groupByKey(row => row.getAs[String]("_1")).flatMapGroups((key, inputItr) => {
        val inputSeq = inputItr.toSeq

        val length = inputSeq.size
        var listBuff = new ListBuffer[Row]()
        var counter : Int= 0
        for(i <- 0 until(length))
        {
          counter+=1

        }

        for(i <- 0 until length ) {
          var x = inputSeq(i)
          listBuff += Row.fromSeq(x.toSeq ++ Array[Int](counter))
        }
        listBuff.iterator
      })(expr1)

      tranform2.show
    }
}

以下是堆栈跟踪

18/11/21 19:39:03 WARN TaskSetManager: Lost task 144.0 in stage 11.0 (TID 400, localhost, executor driver): java.lang.UnsupportedOperationException: fieldIndex on a Row without schema is undefined.
at org.apache.spark.sql.Row$class.fieldIndex(Row.scala:342)
at org.apache.spark.sql.catalyst.expressions.GenericRow.fieldIndex(rows.scala:166)
at org.apache.spark.sql.Row$class.getAs(Row.scala:333)
at org.apache.spark.sql.catalyst.expressions.GenericRow.getAs(rows.scala:166)
at com.quantuting.sparkutils.main.Test$$anonfun$4.apply(Test.scala:59)
at com.quantuting.sparkutils.main.Test$$anonfun$4.apply(Test.scala:59)
at org.apache.spark.sql.execution.AppendColumnsWithObjectExec$$anonfun$9$$anonfun$apply$3.apply(objects.scala:300)
at org.apache.spark.sql.execution.AppendColumnsWithObjectExec$$anonfun$9$$anonfun$apply$3.apply(objects.scala:298)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:409)
at scala.collection.Iterator$$anon$11.next(Iterator.scala:409)
at org.apache.spark.shuffle.sort.BypassMergeSortShuffleWriter.write(BypassMergeSortShuffleWriter.java:149)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
at org.apache.spark.scheduler.Task.run(Task.scala:109)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:345)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)

如何修复此代码？

Answer 1

通过替换getAs[T]方法的fieldname版本（用于groupByKey的函数）可以避免所报告的问题：

groupByKey(row => row.getAs[String]("_1"))

具有field-position版本：

groupByKey(row => row.getAs[String](fieldIndexMap("_1")))

其中fieldIndexMap将字段名称映射到其相应的字段索引：

val fieldIndexMap = tranform.schema.fieldNames.zipWithIndex.toMap

请注意，您对flatMapGroups的功能可以简化为以下形式：

val tranform2 = tranform.groupByKey(_.getAs[String](fieldIndexMap("_1"))).
  flatMapGroups((key, inputItr) => {
    val inputSeq = inputItr.toSeq
    val length = inputSeq.size
    inputSeq.map(r => Row.fromSeq(r.toSeq :+ length))
  })(expr1)

将原始groupByKey/flatMapGroups方法应用于“ dataFrame”与“ transform”之间的不一致行为显然与方法处理DataFrame与Dataset[Row]的方式有关。

Answer 2

JIRA在Spark项目https://issues.apache.org/jira/browse/SPARK-26436上收到的解决方案

此问题是由您如何创建行引起的：

listBuff += Row.fromSeq(x.toSeq ++ Array[Int](counter))

Row.fromSeq创建GenericRow，而GenericRow的fieldIndex未实现，因为GenericRow没有架构。

更改该行以创建GenericRowWithSchema可以解决该问题：

listBuff += new GenericRowWithSchema((x.toSeq ++ Array[Int](counter)).toArray, newSchema)

未定义模式的行上的java.lang.UnsupportedOperationExceptionfieldfieldIndex未定义：row.getAs [String]上的异常

2 个答案: