SEZ9 commented on issue #12879: URL: https://github.com/apache/seatunnel/issues/12879#issuecomment-6258528024
Thanks for the detailed report and the reproducible test cases. The behavior you describe is consistent with what you observed: a nested `ROW` is emitted as a bare record rather than a nullable union, so a null `ROW` field, a null element of `ARRAY<ROW<...>>`, or a null value of `MAP<STRING, ROW<...>>` has no legal encoding. Your compatibility concern is also valid: changing the ROW branch to a union alters the binary encoding for non-null records too, so existing unframed Avro payloads could be affected when the old writer schema is unavailable. Since you are willing to submit a PR, could you open one that adds your three round-trip cases to `AvroSerializationSchemaTest` along with a proposed fix? Including the old-writer/new-reader compatibility cases you mentioned would help decide between an opt-in nullable-ROW mode and a documented default change. Thanks! <!-- streview-comment:1858 --> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
