BUG/API: Consistency in .where() when setting with None for both inplace in a Series (GH7939)

[jreback]

closes #7939

[jorisvandenbossche]

@jreback not fully clear, but which of the two behaviours is it after this PR? The inplace (convert to NaN) or the non-inplace (keep None) ?

[jreback]

I closed this because not really sure what (if anything) to do at this point. You can see the reference discussion on the issue. None is kind of a mess. But changing anything is extremely tricky as their really aren't enough tests to validate current or proposed behavior. and it is often ambiguous on what to do.

np.nan is the missing value indicator, so None should NOT be treated specially, but it IS in some places for convience. It is a can of worms.

[jorisvandenbossche]

ah, yes, I thought this was merged, my mistake.
I know it is a mess, I just added some more of it :-) (#7939 (comment))