在hadoop上的一个流作业中使用多个映射器输入？

Question

In java I would use:

MultipleInputs.addInputPath(conf, path, inputFormatClass, mapperClass)

to add multiple inputs with a different mapper for each.

Now I am using python to write a streaming job in hadoop, can a similiar job be done?

Answer 1

您可以使用多个-input选项来指定多个输入路径：

hadoop jar hadoop-streaming.jar -input foo.txt -input bar.txt ...

Answer 2

I suppose this can help you: https://github.com/hyonaldo/hadoop-multiple-streaming .

Here you can see "different mappers for these different input path" as well:

hadoop jar hadoop-multiple-streaming.jar \  
  -input    myInputDirs \  
  -multiple "outputDir1|mypackage.Mapper1|mypackage.Reducer1" \  
  -multiple "outputDir2|mapper2.sh|reducer2.sh" \  
  -multiple "outputDir3|mapper3.py|reducer3.py" \  
  -multiple "outputDir4|/bin/cat|/bin/wc" \  
  -libjars  "libDir/mypackage.jar" \
  -file     "libDir/mapper2.sh" \  
  -file     "libDir/mapper3.py" \  
  -file     "libDir/reducer2.sh" \  
  -file     "libDir/reducer3.py"