scovich commented on PR #10015: URL: https://github.com/apache/arrow-rs/pull/10015#issuecomment-4624174451
> AFAIK this tweak (hack?) only applies to top-level fields of a shredded variant column. We have no testing for nested uuid columns, but I'm pretty sure they would still write out to parquet as bare binary arrays? Probably not even validated (because `VariantArray::from_parts` does not invoke `canonicalize_shredded_types`)? Thinking more about this -- should `canonicalize_and_verify_data_type` add the UUID extension type to FixedSizeBinary(16) fields it encounters? (it already has a match arm that enforces the length). In theory it would be easier to plumb the necessary Field information through that private helper, but I'm not sure how it would work with top-level UUID fields? Also -- why is `VariantArray::from_parts` the correct (and only) place we should be imposing the UUID extension type? Do other code paths need similar treatment so the behavior is consistent? -- 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]
