Update to Arrow 35.0.0 #5441
Update to Arrow 35.0.0 #5441
Conversation
| @@ -2979,10 +2979,6 @@ mod tests { | |||
| ScalarValue::Decimal128(None, 10, 2), | |||
| ScalarValue::try_from_array(&array, 3).unwrap() | |||
| ); | |||
| assert_eq!( | |||
There was a problem hiding this comment.
This test was wrong, index 4 is beyond the bounds of the array. Previously this was fine as the null buffer only enforced the length for multiples of 8
| "| [] |", | ||
| "| [] |", | ||
| "| [] |", | ||
| "| [a] |", |
| @@ -1411,14 +1411,14 @@ async fn cast_timestamp_before_1970() -> Result<()> { | |||
|
|
|||
| assert_batches_eq!(expected, &actual); | |||
|
|
|||
| let sql = "select cast('1969-01-01T00:00:00.1Z' as timestamp);"; | |||
| let sql = "select cast('1969-01-01T00:00:00.100Z' as timestamp);"; | |||
There was a problem hiding this comment.
what happens without this change?
There was a problem hiding this comment.
An error, it was a mis-reading of the RFC spec, I thought it restricted to multiples of 3 for subsecond precision, I'm working on a fix. It is somewhat amusing that chrono doesn't actually have any tests for this...
There was a problem hiding this comment.
Filed upstream issue: apache/arrow-rs#3859 (I think apache/arrow-rs#3858 fixes it)
There was a problem hiding this comment.
I would tend to think this regression should block the upgrade of datafusion, but would defer to others.
maybe we can mark this PR as "ready for review" to get some other opinions 🤔
| @@ -414,7 +414,7 @@ async fn set_time_zone_bad_time_zone_format() { | |||
| .await | |||
| .unwrap(); | |||
| let err = pretty_format_batches(&result).err().unwrap().to_string(); | |||
| assert_eq!(err, "Parser error: Invalid timezone \"+08:00:00\": Expected format [+-]XX:XX, [+-]XX, or [+-]XXXX"); | |||
| assert_eq!(err, "Parser error: Invalid timezone \"+08:00:00\": only offset based timezones supported without chrono-tz feature"); | |||
There was a problem hiding this comment.
perhaps we can improve the error message a bit in a future 😊
| @@ -431,7 +431,7 @@ mod tests { | |||
| .project(proj)? | |||
| .build()?; | |||
|
|
|||
| let expected = "Error parsing 'I'M NOT A TIMESTAMP' as timestamp"; | |||
| let expected = "Error parsing timestamp from 'I'M NOT A TIMESTAMP'"; | |||
Thanks @comphead -- I think that is apache/arrow-rs#3859 There were a few regressions, in arrow 35.0.0 so I think we should probably skip arrow 35.0.0 and wait for arrow 36.0.0: apache/arrow-rs#3889 |
Do we have a list of regression to be fixed in arrow 36.0.0? |
The ones I know of are apache/arrow-rs#3889 (comment) |
|
Closing in favor of #5685 |
Which issue does this PR close?
Closes #.
Rationale for this change
Update to Arrow 35.0.0
What changes are included in this PR?
Are these changes tested?
Are there any user-facing changes?