jackylee-ch commented on PR #10173: URL: https://github.com/apache/paimon/pull/10173#issuecomment-5953876219
Fixed both P1s in 182d60660. The truncate(N) value path now omits min/max (keeping the null count) when Arrow min/max can't give a sound bound, instead of publishing one: - a float column with any NaN (including all-NaN) records no numeric bound, so a `= NaN` scan isn't wrongly pruned; - a `timestamp[ns]` column records no bound (a Python datetime drops the sub-microsecond part the manifest can't carry), so TIMESTAMP(9) files no longer publish a too-narrow bound. Detection is value-stats-only (key stats stay exact; `full` untouched, its pre-existing behavior is out of scope), and the multi-batch writer ORs the flag across batches. Verified in `metadata_stats_mode_test` (NaN / all-NaN / ns + a full-mode control); neutralizing the omission makes the three fail. flake8 and `git diff --check` clean. -- 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]
