Using mapping of keys should be limited only to CSV

[pombredanne]

The mapping is a convenience that should be strictly limited to CSV output and ingestion .
This should not be a feature whne dealing with JSON.

[pombredanne]

Actually I suggest an even more radical change: We should decouple mapping (as a pure CSV to CSV operation) from the rest of the processing and commands
The new proposed flow would be that IF you want to apply mappings and have some mappings , THEN you will have to run a first command on your CSV that will create a new "mapped" CSV using these mappings. Afterwards the processing is the same.

When using about gen ALL the transformed CSV columns would then be added to the ABOUT files (ignoring "technical" columns as this happens today)

The mapping command would be limited to cut unknown/unmapped columns and rename columns.

The benefit would be vastly simplified code that will be easier to evolve.

[pombredanne]

The gen --mapping-output option is not well defined and should be removed too IMHO

[pombredanne]

This has been implemented and merged