Add to_char function implementation using chrono formats
#9181
Conversation
github-actions
bot
added
logical-expr
|
Thanks @Omega359 -- I hope to get a chance to review this over the weekend |
There was a problem hiding this comment.
Thank you @Omega359 -- like your previous PRs I found this one very well commented, well tested and a joy to review.
I have one question about the handling of scalar / array arguments, which I think a few more tests could help with, but otherwise this PR looks good to me
Thanks again
alamb
changed the title
to_char function implementation using chrono formats
Co-authored-by: Andrew Lamb <andrew@nerdnetworks.org>
…r the Duration type (which is now an alias for TimeDelta) and this broke the timestamp tests.
…re/to_char_9147 # Conflicts: # datafusion/physical-expr/src/datetime_expressions.rs
| Some(date.timestamp_nanos_opt().unwrap()), | ||
| None, | ||
| ), | ||
| StringArray::from(vec!["%Y::%m::%d %S::%M::%H %f".to_string()]), |
| query error DataFusion error: Execution error: Cast error: Format error | ||
| SELECT to_char('2000-02-03'::date, '%X%K'); | ||
|
|
||
| query error DataFusion error: Arrow error: Invalid argument error: column types must match schema types, expected Utf8 but found Null at column index 0 |
There was a problem hiding this comment.
this is probably something we could fix in a follow on PR, as a usability papercut. Might be a good first project for someone 🤔
There was a problem hiding this comment.
I probably should handle it appropriately now to be honest. Originally I think this just defaulted to the default formatting for the type which to me I think is actually the best of the potential solutions (vs almost the worst now).
…efault format being used.
| @@ -2657,16 +2657,19 @@ PT123456S | |||
| query T | |||
| select to_char(arrow_cast(123456, 'Duration(Second)'), null); | |||
| ---- | |||
| NULL | |||
| PT123456S | |||
There was a problem hiding this comment.
Postgres seems to think this should be null
postgres=# select to_char(interval '1 day', null);
to_char
---------
(1 row)
Which issue does this PR close?
Closes #9147
Rationale for this change
The ability to format dates, timestamps or durations is a very useful feature.
What changes are included in this PR?
Code, tests, benchmarks, documentation.
Are these changes tested?
Yes
Are there any user-facing changes?
A new function 'to_char' (alias date_format) is available.