Reverse operators on integer-NA series and numpy scalars resulting in object dtype

[jorisvandenbossche]

This is not necessarily a bug, but at least some unexpected results:

In [83]: s = pd.Series([1, 2, 3], dtype='Int64')

In [84]: s + s[0]
Out[84]: 
0    2
1    3
2    4
dtype: Int64

In [85]: s[0] + s
Out[85]: 
0    2
1    3
2    4
dtype: object

In the above s[0] is a numpy scalar, if you do a radd with a normal int, you do not get object:

In [86]: 1 + s
Out[86]: 
0    2
1    3
2    4
dtype: Int64

In [87]: type(s[0])
Out[87]: numpy.int64

[peterpanmj]

I would like to work on a PR.

[jreback]

go for it!

[peterpanmj]

The issue is resolved by some other PR. In current master it behaves just like expected. @jorisvandenbossche @jreback

In [2]: import pandas as pd

In [3]: s = pd.Series([1, 2, 3], dtype='Int64')

In [4]: s[0]
Out[4]: 1

In [5]: s[0] + s
Out[5]:
0    2
1    3
2    4
dtype: Int64

In [6]: s + s[0]
Out[6]:
0    2
1    3
2    4
dtype: Int64

[jreback]

lots of work by @jbrockmendel and others in ops likely fixed this
i believe we have good test coverage for this but pls take a look
if we do then can just close this

[TomAugspurger]

We have #22974 for testing ops between EA and ndarrays / sequences directly (we just do series right now). So we can close this in favor of the issue for creating a base test.