+1 (non-binding) On Thu, Jun 11, 2026 at 5:22 PM Andrei Tserakhau via dev <[email protected]> wrote: > > Hi all, > > Following the discussion thread [1], I'd like to call a vote on the > spec clarification in PR #16446 [2]. > > Summary of the change: > - The `day` partition transform's result type in the transform table is > changed from `int` to `date`, matching the long-standing default > behavior of Java, PyIceberg, Rust, and C++ (all of which write the > Avro `date` logical type in manifests). > - A note is added clarifying that readers must also accept plain `int` > values for the `day` transform, interpreting each as a date in days > since 1970-01-01. (Reader tolerance is also inherited from the Avro > spec, which requires readers to ignore unrecognized logical-type > annotations.) > > This is a clarification, not a behavioral change: the physical > representation (a 4-byte day count) is unchanged, and no existing > manifests are affected. It documents what implementations already do > and resolves an ambiguity that has been re-litigated several times > (issue #16414 [3], and most recently a real interop failure in > iceberg-go #1176 [4]). > > Per the contributor guide, changes under format/ require a vote even > for clarifications, hence this thread. > > The vote will be open for at least 72 hours. > > [ ] +1 Approve the spec clarification > [ ] +0 > [ ] -1 Do not approve (please provide reasons) > > [1] https://lists.apache.org/thread/wnhvdmwc3b1dxc2ogk7po682pky12cps > [2] https://github.com/apache/iceberg/pull/16446 > [3] https://github.com/apache/iceberg/issues/16414 > [4] https://github.com/apache/iceberg-go/pull/1176 > > Here's my +1 (non-binding). > > Best, > Andrei
-- Regards Junwang Zhao
