BUG: MultiIndex.set_names loses name

[TomAugspurger]

Seems surprising, I thought .name would come along.

In [31]: idx = pd.MultiIndex.from_product([['A'], ['a', 'b', 'c']])

In [32]: idx.name = 'foo'

In [33]: idx2 = idx.set_names(['l1', 'l2'])

In [34]: idx2.name

Bug, or user error since MultiIndex.name isn't all that useful?

[jreback]

I think the fix is similar to what I am doing in RangeIndex for copy, see here: https://github.com/pydata/pandas/pull/11892/files#diff-dc96e62847cdfe77b570cf9c6b0e8086R495

name and names are accepted and aliased to each other for both Index and MultiIndex. mainly for compat I think (e.g. for an Index people tend to use name and for a MultiIndex names). You can't have both, but either should set correctly. (of course a MultiIndex index requires a list equal in len to the levels; .set_names enforces this). I think the setter needs to be updated though.

[jreback]

does #11980 close this?

[TomAugspurger]

No, I think I had an old test would have failed, but I removed the ability to set both self.index.name and self.index.names on a MultiIndex

[jreback]

thanks @TomAugspurger