REF: split out dtype-finding in concat_compat #53260
Conversation
| """ | ||
| from pandas.core.construction import ensure_wrapped_if_datetimelike | ||
|
|
||
| to_concat = [ensure_wrapped_if_datetimelike(x) for x in to_concat] |
There was a problem hiding this comment.
also, could you explain why ensure_wrapped_if_datetimelike and atleast_2d are no longer needed here?
in general, for anything non-trivial, I'd really appreciate a comment explaining why you're making changes, else reviews can really take hours
There was a problem hiding this comment.
the not single_dtype case is irrelevant bc in that case we find_common_dtype and cast to it. The ensure_wrapped_if_datetimelike is unnecessary bc caller is responsible for ensuring we are wrapped where appropriate. The axis is unnecessary bc the axis=1 cases all go through the lib.dtypes_all_equal path at the top of concat_compat
There was a problem hiding this comment.
The axis is unnecessary bc the axis=1 cases all go through the lib.dtypes_all_equal path at the top of concat_compat
Not sure I follow, sorry - in test_append_new_columns, it reaches concat_compat with axis=1 but it doesn't go through the # fastpath! path at the top of concat_compat
There was a problem hiding this comment.
test_append_new_columns doesnt have any datetime columns, so wouldn't go through _concat_datetime. Should have been clearer thats what i was referring to.
|
Thanks @jbrockmendel |
* REF: split out dtype-finding in concat_compat * mypy fixup * fix annotation * remove unused ignore
* REF: split out dtype-finding in concat_compat * mypy fixup * fix annotation * remove unused ignore
Split off from #52532