The clarification is simple and clear from the writer's perspective. CMIW, the implication is that reader should drop bbox with any NaN value regardless of the coordinate axis (in case of a writer bug).
On Wed, May 7, 2025 at 6:21 AM Huang-Hsiang Cheng <hua...@apple.com.invalid> wrote: > +1 (non-binding) > > On May 6, 2025, at 2:53 PM, Ryan Blue <rdb...@gmail.com> wrote: > > +1 (binding) > > Thanks to Jia and Szehon for the quick turn-around getting this done! > > On Tue, May 6, 2025 at 2:37 PM Jia Yu <ji...@apache.org> wrote: > >> +1 (non-binding) >> >> Thanks for putting this together! >> >> Jia Yu >> >> On Tue, May 6, 2025 at 2:09 PM Szehon Ho <szehon.apa...@gmail.com> wrote: >> >>> Hi everyone, >>> >>> As discussed briefly in >>> https://lists.apache.org/thread/ncj0xjh2ct5xvovn4tzc45lkm1wbmorq, there >>> is a minor clarification for geo type bounds that we want to get in for >>> finalizing V3 spec. We want to clarify the behavior of null/NaN coordinate >>> values in geo objects. There can be many coordinate values in a geo >>> object, only the null/NaN ones are skipped, the rest are not skipped. >>> >>> I would like to raise a vote for this minor change to V3 spec. The PR >>> is: https://github.com/apache/iceberg/pull/12956. >>> >>> This is going on in parallel with a parquet-format pr: >>> https://github.com/apache/parquet-format/pull/494 >>> >>> This vote will be open for at least 72 hours. >>> >>> [ ] +1 Add these simplifications to the V3 Geo spec >>> [ ] +0 >>> [ ] -1 I have questions and/or concerns >>> >>> Thanks, >>> Szehon >>> >> >
