Abe encouraged me to discuss this topic in the dev@ thread, Hence the email. I am looking for pointers on this topic.
Currently the ordering of the output generated by the MR engine is dependent on the IDF text format,i,e the CSV text representation. Is this intentional? We now have JSONIDF and AvroIDF in sqoop2 and it would be convenient if the ordering can be custom defined. More details are in this ticket and my proposal is to allow this ordering to be configurable per IDF implementation. https://issues.apache.org/jira/browse/SQOOP-1936 Thoughts? Best, *./Vee*
