TST/BUG (string dtype): Fix and adjust indexes string tests #59544
Conversation
pandas/tests/indexes/test_base.py
Outdated
| if using_infer_string and HAS_PYARROW: | ||
| tm.assert_extension_array_equal( | ||
| new_index.values, pd.array(arr, dtype="str") | ||
| ) |
There was a problem hiding this comment.
This one I don't understand. Why is index.values a numpy array in case of "python" storage?
There was a problem hiding this comment.
Yeah I thought I rolled that one back, the change was just something I tried.
Updated now
pandas/tests/indexes/test_base.py
Outdated
| elif index.dtype == "str" and not index.dtype.storage == "python": | ||
| with pytest.raises(NotImplementedError, match="i8"): | ||
| index.view("i8") |
There was a problem hiding this comment.
This is just testing a different error, right? (just to understand why this has a separate branch)
(ideally we should also make it a TypeError, I assume, maybe can add a TODO note)
| # non-EA dtype indexes have special casting logic, so we punt here | ||
| pass | ||
| if isinstance(data, (set, frozenset)): | ||
| data = list(data) |
There was a problem hiding this comment.
This is related to what you mentioned on slack, I assume?
idx = pd.Index({1, 2}, dtype="Int64")
idx = pd.Index({1, 2}, dtype="int64")
Whether we consider a set as sequence as valid input for constructors or not (IMO it would also be fine to not do that more generally)
There was a problem hiding this comment.
Yes, correct. It#s explicitly allowed for Index, so we shouldn't break this now
There was a problem hiding this comment.
I see, for Series(..) we disallow it explicitly, but for Index(..) it currently works for non-EA dtypes. Sounds good to allow that for now.
|
Ah, still some mypy errors .. |
|
Thanks @phofl |
…ev#59544) Co-authored-by: Joris Van den Bossche <jorisvandenbossche@gmail.com>
…ev#59544) Co-authored-by: Joris Van den Bossche <jorisvandenbossche@gmail.com>
Co-authored-by: Joris Van den Bossche <jorisvandenbossche@gmail.com>
xref #54792