Hadoop-MultipleInputs/MultipleOutputs1<转>

一个Job里可以从多个同质或异质的输入源读取数据，并使用各自的Mapper 

 MultipleInputs.addInputPath(conf, ncdcInputPath,
     TextInputFormat.class, MaxTemperatureMapper.class)
 MultipleInputs.addInputPath(conf, metOfficeInputPath,
     TextInputFormat.class, MetOfficeMaxTemperatureMapper.class);

MultiOutputFormat可以让你按一定规则指定、分隔reduce output的文件名，如 

 ...
 static class StationNameMultipleTextOutputFormat
     extends MultipleTextOutputFormat<NullWritable, Text> {

     private NcdcRecordParser parser = new NcdcRecordParser();

     protected String generateFileNameForKeyValue(NullWritable key, Text value,
         String name) {
       parser.parse(value);
       return parser.getStationId();
     }
   }
 ...