jorisvandenbossche commented on pull request #10176:
URL: https://github.com/apache/arrow/pull/10176#issuecomment-854550555
There seems to be something wrong with the ISO calendar (year)
implementation. I was playing a bit with this branch, and seeing:
```python
In [25]: arr = pa.array([pd.Timestamp("2012-01-01 01:02:03")],
type=pa.timestamp("ns"))
In [26]: pc.iso_calendar(arr)
Out[26]:
<pyarrow.lib.StructArray object at 0x7f5bbf3639a0>
-- is_valid: all not null
-- child 0 type: int64
[
2012 # <-------------- this should be 2011 (since the week
is 52, it falls in the "previous" year)
]
-- child 1 type: int64
[
52
]
-- child 2 type: int64
[
7
]
```
Also in general, for non-nanosecond resolutions, you get wrong results for
all kernels. It seems that it is interpreting the non-nanosecond epoch as
nanosecond epoch. That's related to your previous comment that tests for other
units are still failing?
--
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.
For queries about this service, please contact Infrastructure at:
[email protected]
