hosseinsha commented on PR #25326:
URL: https://github.com/apache/datafusion/pull/25326#issuecomment-6288918838
Some independent numbers on the `ok_or` → `ok_or_else` part of this PR, in
case they help it land:
- In a CPU profile of a production query that groups roughly a year of rows
(about 83 M) by
`date_trunc('month', ts)` in a named session time zone (with DST),
`alloc::fmt::format` and integer
`Display::fmt` under `general_date_trunc` were 8–13 % of the samples — all
of it the eagerly built error.
- With **only** that one-line change on `main` (a7c7c08), the existing
`date_trunc` bench extended with two cases in
`America/New_York` (1,000 nanosecond values) went from 309.3 µs to 163.6
µs for `day` and from 311.4 µs to
165.6 µs for `month`, about −47 %.
Since this PR now conflicts with `main`, would it help if the one-line
`ok_or_else` change landed on its own first
(referencing #25325), leaving the fixed-offset fast path to this PR? Happy
to open that small PR if you and the
maintainers prefer, or to leave it here if this is close to merging.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
