TST: astyp via loc to int64 #36942
TST: astyp via loc to int64 #36942
Conversation
There was a problem hiding this comment.
Thanks @hardikpnsp for the PR!
Looks good! Some small comments
| def test_astype_cast_via_loc_nan_int(self, val, dtype): | ||
| # see GH#31861 | ||
| expected = DataFrame({"a": ["foo"], "b": integer_array([val], dtype=dtype)}) | ||
| result = DataFrame({"a": ["foo"], "b": [val]}) |
There was a problem hiding this comment.
nitpick but i'd call this df. It's not really the result yet since the point of the test is to change it in the next line and then assert
There was a problem hiding this comment.
Thanks for the input! I agree, its better to call it df, I will try to quickly fix these.
| @pytest.mark.parametrize("dtype", ["Int64", "Int32", "Int16"]) | ||
| @pytest.mark.parametrize("val", [np.nan, NA]) | ||
| def test_astype_cast_via_loc_nan_int(self, val, dtype): | ||
| # GH31861 |
There was a problem hiding this comment.
can you add the working conditions as well (e.g. nan + a value) from the OP
There was a problem hiding this comment.
sure! let me try.
| @@ -136,6 +137,15 @@ def test_astype_cast_nan_inf_int(self, val, dtype): | |||
| with pytest.raises(ValueError, match=msg): | |||
| df.astype(dtype) | |||
|
|
|||
| @pytest.mark.parametrize("dtype", ["Int64", "Int32", "Int16"]) | |||
| @pytest.mark.parametrize("val", [np.nan, NA]) | |||
There was a problem hiding this comment.
use nulls_fixture instead and any_nullable_int_dtype fixture
There was a problem hiding this comment.
@jreback , here in nulls_fixture, the test case fails for pd.NaT because it is inferred as datetime64. I think it is expected behavior but would like your opinion on it.
code snippet for reference
>>> df = pd.DataFrame({"a": ["foo"], "b": [pd.NaT]})
>>> df.dtypes
a object
b datetime64[ns]
dtype: object
>>> df.loc[:, "b"] = df.loc[:, "b"].astype('Int64')
TypeError: datetime64[ns] cannot be converted to an IntegerDtypeThere was a problem hiding this comment.
Could use pytest.skipif for the datetime case
or test for the TypeError somehow but I'm not sure that's necessary
jreback
added
Indexing
There was a problem hiding this comment.
Sorry @hardikpnsp for being quiet
Responded to the question below. If you feel like reopening and pushing the changes to parameterization (use fixtures as suggested) this should be good to go in
| @@ -136,6 +137,15 @@ def test_astype_cast_nan_inf_int(self, val, dtype): | |||
| with pytest.raises(ValueError, match=msg): | |||
| df.astype(dtype) | |||
|
|
|||
| @pytest.mark.parametrize("dtype", ["Int64", "Int32", "Int16"]) | |||
| @pytest.mark.parametrize("val", [np.nan, NA]) | |||
There was a problem hiding this comment.
Could use pytest.skipif for the datetime case
or test for the TypeError somehow but I'm not sure that's necessary
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diffAdds a test for a special case of
astypevialocto extension type Int64 when data has np.nan or pd.NA.