REGR: maybe_convert_objects ignoring uints #47475
Conversation
rhshadrach
added
Bug
Dtype Conversions
pandas/_libs/tslibs/util.pxd
Outdated
| """ | ||
| return PyObject_TypeCheck(obj, &PySignedIntegerArrType_Type) | ||
|
|
||
| cdef inline bint is_uinteger_object(object obj) nogil: |
There was a problem hiding this comment.
do you expect these to be used elsewhere? if not, i think the build might be marginally faster (maybe even smaller) if they go directly in lib.pyx
There was a problem hiding this comment.
I'd hazard a guess no. Also, I had reservations on the is_sinteger_object function anyways; it would return False even if e.g. -1 is passed, which is possibly the wrong answer depending on use. But every other function there only looked at the type of object, so it seemed wrong to add a function with value-dependent behavior. Inlining these gets around that.
…16_regr � Conflicts: � doc/source/whatsnew/v1.5.0.rst
| [np.uint64(1)], | ||
| ], | ||
| ) | ||
| def test_constructor_numpy_uints(self, values): |
There was a problem hiding this comment.
are pd.Index or pd.array affected? pd.NumericIndex?
There was a problem hiding this comment.
pd.array and pd.NumericIndex are not impacted; I've added tests for Index.
arr = pd.array([np.uint16(1)])
print(arr)
# <IntegerArray>
# [1]
# Length: 1, dtype: Int64
index = NumericIndex([np.uint16(1)])
print(index)
# NumericIndex([1], dtype='uint16')
Both of these are the same as 1.4.3
|
thanks @rhshadrach |
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.I think this is closely related to #37258, though it does not impact any code examples there.
cc @TomAugspurger, @jbrockmendel, @jorisvandenbossche