CLN: Don't intercept NotImplementedError in _cython_transform #49742
Conversation
pandas/core/groupby/generic.py
Outdated
| @@ -1362,8 +1362,9 @@ def arr_func(bvalues: ArrayLike) -> ArrayLike: | |||
| arr_func, ignore_failures=numeric_only is lib.no_default | |||
| ) | |||
| except NotImplementedError as err: | |||
| # For NotImplementedError, args[0] is the error message | |||
| raise TypeError(err.args[0]) from err | |||
| # For NotImplementedError, args[0] is the error message when available | |||
There was a problem hiding this comment.
im fine with this, but FWIW id just not catch this and let the NotImplementedError bubble up
There was a problem hiding this comment.
Sounds good - I'll take that approach here. The only functions in our test that currently raise here are cumsum, cummin, cummax, and cumprod. I'm thinking longer term it'd be better to provide a Python fallback for object dtypes where cumsum (on strings, at least), cummin, and cummax could be successful, and cumprod will then raise the proper TypeError. Does that make sense?
rhshadrach
changed the title
|
lgtm. is this not user facing at all / worth a whatsnew? |
This was changed from raising NotImplementedError to a TypeError in #49665. So this PR is really just reverting that change. |
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.Ref: #49665 (comment)
The one place in groupby where there is
raise NotImplementedErroris in_reconstruct_ea_result. This is not hit by tests.cc @jbrockmendel