Map and Reduce support corresponding Map/Reduce methods, setup methods, and cleanup methods. The setup method is called before the Map/Reduce method, and each worker calls it only once. 

The cleanup method is called after the map/reduce method, and each worker calls it only once.

For a detailed example, see Program examples.


Some columns in output key records can be taken as sort columns, but user-defined comparator is not supported. You can select several columns from the sort column as Group columns, but the user-defined Group comparator is not supported. Sort columns are used to sort your data while Group columns are used for a Secondary Sort.

For more information, see SecondarySort Example.


Supports setting the partition column and customized partitioner. Partition columns have a higher priority than customized partitioners.

According to Hash logic, the partitioner distributes the output data on the Map terminal to different Reduce Workers.


Combines adjacent records in the Shuffle stage. You can choose whether to use Combiner according to different business logic.

Combiner helps to optimize the MapReduce computing framework and the logic of Combiner is generally similar to Reduce. After Map outputs the data, the framework performs a local combiner operation for the data which has the same key value on the Map terminal.

For more information, see WordCount code examples.