BUG: `.map` & `.apply` would convert element type for extension array.

[colinfang]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

s = pd.Series([1, 2, None], dtype="Int32")
s.apply(lambda x: print(x))
s.map(lambda x: print(x))
# print 1.0 2.0 nan instead of 1 2 <NA>

Issue Description

s has dtype Int32Dtype which can encode nullable integers

However, when we use .apply(f) or .map(f), the element passed to f becomes float, and pd.NA becomes np.nan

Expected Behavior

f should receive unconverted value, as if we are doing f(s.iloc[0]), f(s.iloc[1]), ...

Installed Versions

tested with Pandas 2.2.3

[rhshadrach]

Thanks for the report! This is due to #55058, but I don't see much discussion there nor issues that change closes. It does seem surprising to me that one gets np.nan here. If we remove the method on BaseMaskedArray entirely, I see expected behavior (at least in this case). That includes the use of apply (which uses map internally).

cc @WillAyd @mroeschke @phofl

[WillAyd]

Agreed this looks strange - the function should be receiving the arguments as integers / pd.NA

[KevsterAmp]

take

[KevsterAmp]

@rhshadrach - What method are you referring on BaseMaskedArray to be removed? Thanks