BUG: Series.min() fails for string series containing None

[miketkelly]

>>> s = pd.Series([None, 'a', 'b'])
>>> s
0    None
1       a
2       b
dtype: object
>>> s.min()
inf

[jreback]

this is an undefined operation

what would you have this return?

[miketkelly]

In Python 2 at least the object ordering is well defined. I'd expect similar behavior to numpy, although I'd prefer 'a' as an answer. I just don't get inf.

>>> pd.np.min(s.values) == None
True

Is this an undefined operation because its an object series? Or is it undefined because it as None in it?

>>> s = pd.Series([None, Decimal(1), Decimal(2)])
>>> s
0    None
1       1
2       2
dtype: object
>>> s.min()
Decimal('1')
>>> s.max()
Decimal('2')
>>> pd.np.min(s.values) == None
True

[jreback]

None are really a funny object, as in any other dtype they are converted to nan. It should be treated as missing though, so marking as a bug.

[jreback]

dupe of #4147