BUG: remove usage of _constructor._get_axis_number (fix when _constructor properties return callables) #46018
Conversation
…ctor properties return callables)
|
Found one more, in |
|
tests are failing as well :-> |
|
Ah, that's another case, but where we didn't call a method on it, but used in an |
pandas/core/reshape/concat.py
Outdated
| @@ -539,7 +542,7 @@ def __init__( | |||
| self.new_axes = self._get_new_axes() | |||
|
|
|||
| def get_result(self): | |||
| cons: type[DataFrame | Series] | |||
| cons: Callable | |||
There was a problem hiding this comment.
could annotate as Callable returning DataFrame|Series
There was a problem hiding this comment.
The Callable typing is actually giving errors .. (with or without return type)
There was a problem hiding this comment.
Ah, but I probably have to update the actual properties' annotation to callable as well
|
Are people OK with considering this for 1.4.x? (the actual change is quite small, the bulk of the diff is due to typing) This is actually fixing a regression for geopandas, although not a regression in 1.4.0, but somewhere between 1.0 and 1.4. |
|
@meeseeksdev backport 1.4.x |
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
This comment was marked as resolved.
|
@jorisvandenbossche if you can push a manual backport |
…axis_number (fix when _constructor properties return callables)
|
The auto backport failed as #45272 is not backported. The changes here to that code were only the mypy comments. Those changes have therefore been omitted in the manual backport. |
…r (fix when _constructor properties return callables) (#46228) Co-authored-by: Joris Van den Bossche <[email protected]>
…ctor properties return callables) (pandas-dev#46018)
See report at #32860 (comment) (and we also run into this one case in geopandas). Also related to #32638
As far as I know this is the only remaining case where we rely on
_constructor/_constructor_sliced/_constructor_expanddimreturning a class and not a generic callable (in geopandas, we also have a workaround for this one case).This specific case was introduced in #34837 to avoid circular dependencies/imports, so therefore I added the DataFrame import inline.