bjchambers commented on a change in pull request #849:
URL: https://github.com/apache/arrow-rs/pull/849#discussion_r735126070
##########
File path: arrow/src/csv/writer.rs
##########
@@ -241,7 +241,7 @@ impl<W: Write> Writer<W> {
.value_as_datetime(row_index)
.unwrap(),
};
- format!("{}", datetime.format(&self.timestamp_format))
+ format!("{}Z", datetime.format(&self.timestamp_format))
Review comment:
I think either is an improvement. I believe what I observed when filling
the bug was that the interpretation of the csv after a write+read cycle
depended on the system time zone.
It sounds like any of these proposed approaches will be a short term
improvement (for testing / predictability) and can be further improved with
better time zone support later.
I do think there are cases that will come up not well handled by the current
spec such as per-row timelines (when handling events from multiple users in the
users timeline for instance). But I think that can be handled at least in part
by normalizing on utc for math and having a print kernel that takes timestamp
and time zone columns and prints each timestamp in the given time zone. But
that would also eventually affect the calendar math.
All that to say - this is an improvement and there is still probably work to
be done.
--
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]
