Feature/nw datediffleapyeartransformer #394
Conversation
There was a problem hiding this comment.
thanks for this @ChristinaKatsiafa. Think there is still a bit to do, as I think you need to also narwhalify calculate_age so we can remove the pandas specific logic in transform.
tubular/dates.py
Outdated
|
|
||
| X[self.new_column_name] = X.apply(self.calculate_age, axis=1) | ||
| X = X.to_native() |
There was a problem hiding this comment.
Q: not sure I get why we are doing this last step, what dtype is coming out of the above operation? Understand we were getting issues with the dtype in the expected dataframes for tests when nulls present but did not realise we were also casting here.
There was a problem hiding this comment.
This is because narwhals returns the results as int64[pyarrow] instead of Int64. So I just force the change.
There was a problem hiding this comment.
I'm not sure returning int64[pyarrow] is a problem so would be inclined to leave it and update the test rather than here. What do you think?
There was a problem hiding this comment.
The change is ready to review :)
This is a pull request to convert datediffleapyeartransformer to narwhals.