scala - 使用 MRUnit 1.1.0 测试多个输出

Question

相关问题@Testing multiple outputs with MRUnit但答案不适用于较新的版本 1.1.0

问题是如何设置多个命名输出，以便底层模拟实现识别命名路径。我写信是为了将相同的 reducer 记录写入 2 条路径。我可以通过调用 MultipleOutputs.addNamedOutput(job, "mos", ...) 在常规 MR 作业中做同样的事情

当我尝试运行 mrunit 时，出现以下异常

Named output 'mos' not defined
java.lang.IllegalArgumentException: Named output 'mos' not defined
  at org.apache.hadoop.mapreduce.lib.output.MultipleOutputs.checkNamedOutputName(MultipleOutputs.java:256)
  at org.apache.hadoop.mapreduce.lib.output.MultipleOutputs.write(MultipleOutputs.java:426)
  at TestMultipleOutputsAction$TestReducer$$anonfun$reduce$1.apply(TestMultipleOutputs.scala:48)
  at TestMultipleOutputsAction$TestReducer$$anonfun$reduce$1.apply(TestMultipleOutputs.scala:47)
  at scala.collection.Iterator$class.foreach(Iterator.scala:727)
  at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
  at scala.collection.IterableLike$class.foreach(IterableLike.scala:72)
  at scala.collection.AbstractIterable.foreach(Iterable.scala:54)
  at TestMultipleOutputsAction$TestReducer.reduce(TestMultipleOutputs.scala:47)
  at TestMultipleOutputsAction$TestReducer.reduce(TestMultipleOutputs.scala:35)

scala代码贴在这里。为代码的冗长道歉。我尝试拉入所有部分，以便于独立运行代码。

import org.apache.hadoop.conf.Configuration
import org.apache.hadoop.mrunit.mapreduce.MapReduceDriver
import org.apache.hadoop.io._
import org.apache.hadoop.mapreduce.{Counters, TaskInputOutputContext, Reducer, Mapper}
import org.apache.hadoop.mapreduce.lib.output.MultipleOutputs
import scala.collection.JavaConversions._
import scala.collection.JavaConverters._
import org.scalatest.FunSuite
import org.apache.hadoop.io.SequenceFile.{Writer, Reader}
import java.nio.file.{Path, Paths, Files}
import com.typesafe.scalalogging.slf4j.Logging
import org.apache.hadoop.fs.{Path => HadoopFSPath}

object TestMultipleOutputsAction {
  class TestMapper extends Mapper[LongWritable, MapWritable, LongWritable, MapWritable] with Logging {
    override def setup(context: Mapper[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
    }

    override def cleanup(context: Mapper[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
    }

    override def map(key: LongWritable, value: MapWritable, context: Mapper[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
      context.write(key, value)
    }
  }

  class TestReducer extends Reducer[LongWritable, MapWritable, LongWritable, MapWritable] with Logging {
    var multipleOutputs: MultipleOutputs[LongWritable, MapWritable] = null

    override def setup(context: Reducer[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
      multipleOutputs = new MultipleOutputs[LongWritable, MapWritable](context.asInstanceOf[TaskInputOutputContext[_, _, LongWritable, MapWritable]])
      super.setup(context)
    }

    override def cleanup(context: Reducer[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
    }

    override def reduce(key: LongWritable, values: java.lang.Iterable[MapWritable], context: Reducer[LongWritable, MapWritable, LongWritable, MapWritable]#Context) {
      values.foreach(value => {
        multipleOutputs.write("mos", key, value, "outputPath1")
        multipleOutputs.write("mos", key, value, "outputPath2")
      })
    }
  }
}

object TestHelper extends Logging {
  def generateInput(conf: Configuration, deleteOnExit: Boolean): String = {
    val dirPath = Files.createTempDirectory(Paths.get("/tmp"), "multiple_outputs")

    val filePath = Files.createTempFile(dirPath, "part-m-", ".0001")

    if (deleteOnExit) {
      filePath.toFile.deleteOnExit()
    }

    logger.info(s"writing to path [${filePath.toFile.getAbsolutePath}] ...")

    val seqFilePath = new HadoopFSPath(filePath.toFile.getAbsolutePath)
    val writer = SequenceFile.createWriter(conf,
      Writer.file(seqFilePath), Writer.keyClass(classOf[LongWritable]),
      Writer.valueClass(classOf[MapWritable]))

    for (i <- 1 to 10) {
      val mapWritable = new MapWritable()
      mapWritable.put(new Text("mod2"), new LongWritable(i % 2))

      writer.append(new LongWritable(i), mapWritable)
    }

    writer.close()

    logger.info(s"writing to path [${filePath.toFile.getAbsolutePath}] completed")

    dirPath.toFile.getAbsolutePath
  }

  def readInput(conf: Configuration
                , path: String
                , mapReduceDriver: MapReduceDriver[LongWritable, MapWritable, LongWritable, MapWritable, LongWritable, MapWritable]) {
    val entries = Files.newDirectoryStream(Paths.get(path), "part-m-*")
    var numRecords = 0

    entries.asScala.foreach(entry => {
      val entryName = entry.toFile.getName
      val absolutePath = entry.toFile.getAbsolutePath

      logger.debug(s"entry name : [${entryName}], absolute path : [${absolutePath}]")

      val validEntry = entryName.startsWith("part-m-")

      if (validEntry) {
        logger.debug(s"adding inputs from path : [${absolutePath}] ...")

        val hadoopPath = new HadoopFSPath(absolutePath)
        val reader = new SequenceFile.Reader(conf, Reader.file(hadoopPath))

        var key = new LongWritable()
        var mapWritable = new MapWritable()
        var numFileRecords = 0

        while (reader.next(key, mapWritable)) {
          logger.debug(key + "\t" + mapWritable)

          mapReduceDriver.addInput(key, mapWritable)

          numFileRecords = numFileRecords + 1
          numRecords = numRecords + 1
        }

        logger.debug(s"adding inputs from path : [${absolutePath}] completed. num file records : [${numFileRecords}]")
      }
    })

    logger.debug(s"adding inputs from path : [${path}] completed. num records : [${numRecords}]")
  }

  def writeOutput(conf: Configuration, dirPath: Path, outputPairs: java.util.List[org.apache.hadoop.mrunit.types.Pair[LongWritable, MapWritable]], deleteOnExit: Boolean): Unit = {
    val filePath = Files.createTempFile(dirPath, "part-m-", ".0001")

    if (deleteOnExit) {
      filePath.toFile.deleteOnExit()
    }

    logger.info(s"writing to path [${filePath.toFile.getAbsolutePath}] ...")

    val seqFilePath = new HadoopFSPath(filePath.toFile.getAbsolutePath)
    val writer = SequenceFile.createWriter(conf,
      Writer.file(seqFilePath), Writer.keyClass(classOf[LongWritable]),
      Writer.valueClass(classOf[MapWritable]))

    outputPairs.asScala.toSeq.foreach(outputPair => {
      logger.debug(s"key : [${outputPair.getFirst}], value : [${outputPair.getSecond}]")
      writer.append(outputPair.getFirst, outputPair.getSecond)
    })

    writer.close()

    logger.info(s"writing to path [${filePath.toFile.getAbsolutePath}] completed")
  }

  def checkCounters(counters: Counters): Unit = {
    counters.getGroupNames.asScala.foreach(groupName => {
      counters.getGroup(groupName).iterator().asScala.foreach(counter => {
        logger.debug(s"groupName: [${groupName}], counterName: [${counter.getName}], counterValue : [${counter.getValue}]")

      })
    })
  }
}

object TestMultipleOutputs extends FunSuite with Logging {
  def testMultipleOutputs(conf: Configuration, inputPath: String, deleteOnExit: Boolean) {
    logger.info(s"TESTINPUT : input path : [${inputPath}] ...")

    val mapReduceDriver = new MapReduceDriver[LongWritable, MapWritable, LongWritable, MapWritable, LongWritable, MapWritable]()
        .withMapper(new TestMultipleOutputsAction.TestMapper)
        .withReducer(new TestMultipleOutputsAction.TestReducer)

    mapReduceDriver.addMultiOutput("mos", classOf[LongWritable], classOf[MapWritable])

    val parentOutputPath = Files.createTempDirectory(Paths.get("/tmp"), "pr_output")

    if (deleteOnExit) {
      parentOutputPath.toFile.deleteOnExit
    }

    TestHelper.readInput(conf, inputPath, mapReduceDriver)

    val outputPairs = mapReduceDriver.run()

    TestHelper.writeOutput(conf, parentOutputPath, outputPairs, deleteOnExit)
    TestHelper.checkCounters(mapReduceDriver.getCounters())

    logger.info(s"TESTINPUT : input path : [${inputPath}] completed")
  }
}

class TestMultipleOutputs extends FunSuite with Logging {
  test("multiple outputs action") {
    val deleteOnExit = true
    val conf = new Configuration()

    val inputPath = TestHelper.generateInput(conf, deleteOnExit)
    TestMultipleOutputs.testMultipleOutputs(conf, inputPath, deleteOnExit)
  }
}

Answer 1

我在 Java 中遇到了同样的问题，并用

注释了我的单元测试

@RunWith(PowerMockRunner.class)
@PrepareForTest(PricePerPlacementReducer.class)

在进行了正确的导入后(基本上是 powermock 版本 1.5.1 及其 junit 绑定(bind)程序)为我解决了这个问题。