CLN: avoid _internal_get_values in pandas._testing #32570
Conversation
| @@ -1038,7 +1038,8 @@ def assert_extension_array_equal( | |||
|
|
|||
| if hasattr(left, "asi8") and type(right) == type(left): | |||
| # Avoid slow object-dtype comparisons | |||
| assert_numpy_array_equal(left.asi8, right.asi8) | |||
| # np.asarray for case where we have a np.MaskedArray | |||
| assert_numpy_array_equal(np.asarray(left.asi8), np.asarray(right.asi8)) | |||
There was a problem hiding this comment.
How can asi8 ever return a MaskedArray?
There was a problem hiding this comment.
we have one test where we pass a MaskedArray to the constructor (I think DTI)
There was a problem hiding this comment.
we have one test where we pass a MaskedArray to the constructor (I think DTI)
But shouldn't that be converted to a normal array upon construction?
There was a problem hiding this comment.
id have no problem with enforcing this, but AFAIK we dont at the moment
| obj=str(obj), | ||
| ) | ||
| elif is_extension_array_dtype(left.dtype) or is_extension_array_dtype(right.dtype): | ||
| assert_extension_array_equal(left._values, right._values) |
There was a problem hiding this comment.
assert_extension_array_equal doesn't work for categorical?
There was a problem hiding this comment.
We have kwargs check_categorical and check_category_ordered and a categorical-specific check at the end of assert_series_equal. xref #32571 trying to cut down on these
jreback
added
Clean
Testing
|
thanks |
| check_dtype=check_dtype, | ||
| obj=str(obj), | ||
| ) | ||
| elif is_extension_array_dtype(left.dtype) or is_extension_array_dtype(right.dtype): |
There was a problem hiding this comment.
Before this was "and", and now it is "or". That seems to cause issues in geopandas in case you specify check_dtype=False (in which case you can eg have an EA dtype and object dtype with equal values)
7 more usages left after this