+1 (non-binding) On 2026/05/19 04:37:16 Micah Kornfield wrote: > +1 (mostly reviewed the spec) as I think Ed also tested to make sure that > old java and c++ bindings won't break with this (sounds like rust versions > pre-footer parsing changes might have issues)
Yes, I can double check, but I believe parquet-rs before 57.0.0 will fail on parsing the metadata when it encounters the new column order (it did when I did the first PoC). I did confirm arrow 20.0.0 had no issues with the file in parquet-testing. Ed > > On Monday, May 18, 2026, Gang Wu <[email protected]> wrote: > > > Hi, > > > > I would like to propose a vote on adopting the format change described > > in PARQUET-2249: IEEE 754 total order & NaN-counts. > > > > The discussion on the dev mailing list can be found here: > > https://lists.apache.org/thread/lzh0dvrvnsy8kvflvl61nfbn6f9js81s > > > > The proposed format specification changes are available in the following > > PR: > > https://github.com/apache/parquet-format/pull/514 > > > > To verify this design's compatibility and correctness, we developed > > two independent PoC implementations: > > 1. Java: https://github.com/apache/parquet-java/pull/3393 > > 2. Rust: https://github.com/apache/arrow-rs/pull/9619 > > > > Both PoCs verified a test file produced by the Parquet-Java PoC: > > https://github.com/apache/parquet-testing/pull/104 > > > > The vote will be open for at least 72 hours. > > > > [ ] +1 Approve the proposed format change > > [ ] +0 No opinion > > [ ] -1 Do not approve (please provide specific reasons) > > > > Thanks, > > Gang Wu > > >
