-
Notifications
You must be signed in to change notification settings - Fork 914
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[REVIEW] Support datetime64
in row-wise op
#6415
Conversation
Sounds like a good idea 🙂 |
Seems EDIT: You also will have to do a |
Thanks @seberg! We don't support |
cc @brandon-b-miller for dtype related discussion since you've been working on implementing cudf dtypes. |
I'll look into if this affects anything and report back - I know we use it in a few places in join. |
Please update the changelog in order to start CI tests. View the gpuCI docs here. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There are two other things I would like to have some comments:
- Memory footprint. I could not come up with an easier solution when there are mixed resolution of
datetime64
, since the underlying values are stored w.r.t the resolution. - (Question thanks to @brandon-b-miller ) What are the expected behavior when overflow happens when converting to a higher resolution?
Skip one testcase due to pandas issue PR 6415
datetime64[ms]
in row-wise opdatetime64[ms]
in row-wise op
- Separating pandas bug cases into separate assertion - Remove excess skipif PR 6415
Address review comments PR6415
Codecov Report
@@ Coverage Diff @@
## branch-0.17 #6415 +/- ##
===============================================
+ Coverage 82.72% 83.11% +0.38%
===============================================
Files 90 90
Lines 14444 14816 +372
===============================================
+ Hits 11949 12314 +365
- Misses 2495 2502 +7
Continue to review full report at Codecov.
|
- Handle both type and value exception for result type - Comment to explain why test memory idendity - Use with pytest.raise block
Addresses review comments: - Using a wrapper of np.find_common_type to accomodate future need of cudf type promotion rules - Avoids the need to modify dataframe by deep copying underlying values used for row-wise op. PR 6415
Co-authored-by: Keith Kraus <[email protected]>
datetime64[ms]
in row-wise opdatetime64
in row-wise op
cc @brandon-b-miller to review again before we merge |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
Closes #6129
In solving this issue a bug from numpy surfaced: numpy/numpy#17428
find_common_types
is not suggested. Should we start replacing the use of this function tonp.promote_types
?