DanielLeens commented on issue #12745: URL: https://github.com/apache/seatunnel/issues/12745#issuecomment-6145190931
Thanks for the careful separation from #12738. The report identifies two coupled pre-existing gaps on current `dev`: `SchemaUtils.convertSchema` creates an Arrow `Map` with no child fields, while `FragmentConverter.writeMapToVector` expects direct key/value children. Canonical Arrow Map fields instead carry an `entries` Struct child containing those fields. Changing only the writer guard would therefore not make newly created datasets work. Yes, please open one focused follow-up PR, separate from the batching work in #12738. It should establish one canonical Map schema contract, have both generated and restored schemas follow it, and make the writer read the key/value fields from the entries Struct. Please cover a canonical restored schema, a generated schema, single- and multi-row values, empty maps, null map rows, nullable values, and read-back values rather than row counts alone. Add the matching English and Chinese connector documentation or type-support notes if the connector's documented contract changes. Please link the PR here when ready. -- 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]
