Xuanwo commented on issue #9183: URL: https://github.com/apache/arrow-rs/issues/9183#issuecomment-6946825706
I can pick this up and drive it to completion. On scope: let's keep this issue focused on swapping the timezone engine behind `arrow_array::timezone::Tz`, which fixes DST not being applied past ~2099, as hit in apache/datafusion-comet#6816. This work does not remove `chrono` from the public API, so we should track public API removal in a separate issue. Four decisions to align on: 1. Move to `jiff` 1.0.0 now (released 2026-11-15, MSRV 1.85, arrow-rs is on 1.88). 2. Use the bundled tzdb via `jiff-tzdb` + `TimeZone::tzif` only, never `jiff`'s global `db()`. 3. Clamp years beyond ±9999 to the nearest covered instant for offset lookups, and narrow kernels from ±262k to ±9999 with a release note. No leap-second fidelity, and return an error when `:60` cannot be represented. 4. Performance gate: within 5% for 1970–2038 and back to noise for 2038–2100. The current 1.7–2.4x slowdown is not acceptable. For the order of work, I'll first take over #11482 (rebase, update to `jiff` 1.0, and clear the 9 open review threads, including `Etc/Unknown`, the duplicated cache entries for different name casings, the bundled-only guarantee, and the CI matrix) so we can merge it and let downstream test the opt-in feature early. Second, I'll open a PR adding a per-zone lazily built transition table to restore 2038–2100 parity, then do the crate-by-crate internal migration with one PR per crate. Switching the default and removing `chrono` are separate later decisions. Three quick checks before I start: - @andygrove: Can I push to `jiff-named-timezones`, or should I open a fresh PR with you as co-author? I prefer pushing to your branch if that works. - @comphead: Are you still on this? If so, happy to split the performance work with you. - @alamb: Are you okay with me rewriting the issue body into this checklist and opening a separate issue for removing `chrono`? -- 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]
