> I think we need to revisit the way this fix works. At this point, it probably 
> makes sense to close the PR and go back to a design discussion first. I would 
> suggest to have that design discussion on the JIRA issue first.
> 
> (1) Where should the BOM be read? Beginning of the file only? Then it should 
> most likely be part of split generation.
> 
> (2) Any design cannot have additional logic in the method that is called per 
> record, otherwise it will have too much performance impact.

@StephanEwen Thank you for your suggestion. I close the PR first. Then discuss 
how to fix this bug.

[ Full content available at: https://github.com/apache/flink/pull/6710 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to