Cudf: [BUG] Comparison based binaryops return nulls instead of True/False in comparing with nulls

[kkraus14]

In [4]: print(cudf.Series([1,2,3,4,5]) == cudf.Series([None, None, None, None, None]))                                                                        
0     
1     
2     
3     
4     
dtype: bool

In [5]: pd.Series([1,2,3,4,5]) == pd.Series([None, None, None, None, None])                                                                                   
Out[5]: 
0    False
1    False
2    False
3    False
4    False
dtype: bool

[kkraus14]

@harrism We can easily handle this at the Python side if you think it's appropriate to do so.

[daxiongshu]

One more example.

>>> pd.Series([None, None, None, None, None]) == pd.Series([None, None, None, None, None])
0    False
1    False
2    False
3    False
4    False
>>> pd.Series([None, None, None, None, None]) != pd.Series([None, None, None, None, None])
0    True
1    True
2    True
3    True
4    True

Basically it should be always True for __ne__ only and always False for __eq__ and every other operator, irrespective of what None compares against.

[harrism]

@kkraus14 is this urgent for 0.9? Assigning @devavret but waiting to hear about urgency before putting it on the 0.9 board.

[revans2]

This is one of the places where SQL and pandas differ.

SELECT foo = NULL from bar

will return one NULL for each row in bar. I personally would prefer to see this be fixed on the python side. We could also have separate operations or separate options for NULL handling in bin-ops.

[kkraus14]

@revans2 Thanks for the feedback! @harrism sounds like it makes sense to handle this on the Python side instead.

[kkraus14]

Closing this because as of Pandas 1.0 we should be propagating nulls as expected.