Github user mattyb149 commented on the issue:
https://github.com/apache/nifi/pull/2430
Reusing the Ambari format certainly makes this one easier to implement, and
as you said the user can convert later with a record processor, but I'm
thinking it might be best to be able to specify the RecordWriter in the
reporting task. That way the conversion, filtering, etc. can be done while the
records are already in object format (i.e. parsed), which would save the step
of re-parsing and re-writing. Using a JsonRecordSetWriter with "Inherit Record
Schema" would result in the same behavior as you propose here, or you could
provide a different format and/or filter out the fields, etc. by using a
configured RecordSetWriter.
@markap14 suggested to encode the "input" schema as a file in
src/main/resources and read it in when the class/instance is created, using the
available util method to convert it to a RecordSchema. To be honest I would
like to have the S2S Provenance reporting task do this as well, but that's a
bit more invasive since it already exists. Since the one in this PR is new,
would be nice to have it be the exemplar for creating reporting services in the
future. What do you think?
---
