jecsand838 commented on PR #8039: URL: https://github.com/apache/arrow-rs/pull/8039#issuecomment-3155960947
@scovich > LGTM. > > Aside: I get the appeal of zero-copy schemas, but I'm pretty sure this schema store will be very difficult to use in practice unless all possible schemas are known up front. Adding a new schema to the store partway through decoding will be ~impossible. But that's a problem with the existing schema API, not this new schema store. Avro pretty much requires you to know all possible schemas upfront. The one inconvenience I can foresee is related to developing a `SchemaStore` trait which can stay up to date with an external registry. However in a real world scenario what would likely occur is the current `reader_schema` becoming another `writer_schema` and a new `reader_schema` being assigned. So this would have more complications than just the lifetimes. I think for this initial implementation it's acceptable to have the caller responsible for making a new `Decoder` upon schema change. Just my 2 cents though of course. -- 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: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
