I'm generally a +1 here since any implementation not using a long would have hit a bug a long time ago when interacting with any of the major engines
I do want to make sure we let this vote go for at least a few more days to tease out any users with strong opinions. Unless we see a major implementation that uses strings pop up soon, I think we should go forward with the patch. On Mon, Jul 28, 2025 at 12:44 PM Steven Wu <[email protected]> wrote: > +1 for fixing the mistake in spec > > On Mon, Jul 28, 2025 at 10:41 AM Steve <[email protected]> wrote: > >> +1 for using long type for snapshotId >> >> On Mon, Jul 28, 2025 at 6:24 AM Péter Váry <[email protected]> >> wrote: >> >>> +1 for long >>> >>> Given that it is implemented as a long in every known implementation, we >>> might not even want to handle the type difference in code >>> >>> Eduard Tudenhöfner <[email protected]> ezt írta (időpont: 2025. >>> júl. 28., H, 12:47): >>> >>>> I agree that this should have been a long in the spec, so +1 to fixing >>>> the spec. I checked and Trino also implements this as a long. >>>> >>>> On Mon, Jul 28, 2025 at 12:39 PM Ajantha Bhat <[email protected]> >>>> wrote: >>>> >>>>> Hi everyone, >>>>> One of the users has raised a PR to update the table >>>>> statistics (puffin stats) spec. >>>>> https://github.com/apache/iceberg/pull/13513 >>>>> >>>>> I have suggested a mailing list voting thread and also tagged the >>>>> original spec author. >>>>> Since there was no response from them for a long time, I am taking it >>>>> forward. >>>>> >>>>> Spec <https://iceberg.apache.org/spec/#table-statistics> mentions the >>>>> snapshot id as String whereas java >>>>> <https://github.com/apache/iceberg/blob/main/api/src/main/java/org/apache/iceberg/StatisticsFile.java#L32> >>>>> and python >>>>> <https://github.com/apache/iceberg-python/blob/479e6639103be367e218c16e83c22bc893400eb3/pyiceberg/table/statistics.py#L35> >>>>> implementations use Long. >>>>> IMO, we can update the implementation to have a string to match the >>>>> spec and handle compatibility during read. >>>>> But the spec is very old and definitely wrong (doesn't align with >>>>> regular snapshot id representation). >>>>> Hence, I think updating the spec is the right option here as current >>>>> implementations like java and python library use long for snapshot id. >>>>> >>>>> Please take a look at the PR and cast your vote. >>>>> >>>>> - Ajantha >>>>> >>>>>
