Split reduced data into output and new input in Hadoop

I've been looking around for days trying to find a way using reduced data for further mapping in hadoop. I've got objects of class A as input data and objects of class B as output data. The Problem is, that while mapping not only Bs are generated but new As as well.

Here's what I'd like to achieve:

1.1 input: a list of As
1.2 map result: for each A a list of new As and a list of Bs is generated
1.3 reduce: filtered Bs are saved as output, filtered As are added to the map jobs

2.1 input: a list of As produced by the first map/reduce
2.2 map result: for each A a list of new As and a list of Bs is generated
2.3 ...

3.1 ...

You should get the basic idea.

I've read a lot about chaining but I'm not sure how to combine ChainReducer and ChainMapper or even if this would be the right approach.

So here's my question: How can I split the mapped data while reducing to save one part as output and the other part as new input data.

Copyright Notice：Content Author:「Mennny」，Reproduced under the CC 4.0 BY-SA copyright license with a link to the original source and this disclaimer.
Link to original article：https://stackoverflow.com/questions/14305351/split-reduced-data-into-output-and-new-input-in-hadoop