Index.name is not validatied to be Hashable

[TomAugspurger]

We require that Series.name be hashable. We don't do the same for Index.name. Should we?

In [4]: pd.Index([], name=[])
Out[4]: Index([], dtype='object', name=[])

In [5]: pd.Series([], name=[])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-5-2c73ddde103e> in <module>
----> 1 pd.Series([], name=[])

~/sandbox/pandas/pandas/core/series.py in __init__(self, data, index, dtype, name, copy, fastpath)
    326         generic.NDFrame.__init__(self, data, fastpath=True)
    327
--> 328         self.name = name
    329         self._set_axis(0, index, fastpath=True)
    330

~/sandbox/pandas/pandas/core/generic.py in __setattr__(self, name, value)
   5257             object.__setattr__(self, name, value)
   5258         elif name in self._metadata:
-> 5259             object.__setattr__(self, name, value)
   5260         else:
   5261             try:

~/sandbox/pandas/pandas/core/series.py in name(self, value)
    468     def name(self, value):
    469         if value is not None and not is_hashable(value):
--> 470             raise TypeError("Series.name must be a hashable type")
    471         object.__setattr__(self, "_name", value)
    472

TypeError: Series.name must be a hashable type

[WillAyd]

Makes sense - I would think generically any labels should have this validation

[R1j1t]

I was looking at this issue and would it be fine if i update this line to

Current version

        if name is None and hasattr(data, "name"):
            name = data.name

Proposed solution

        if name is None and hasattr(data, "name"):
            if not is_hashable(data.name):
                raise TypeError("Index.name must be a hashable type")
            name = data.name

I am happy to raise a PR after adding appropriate tests.

any labels should have this validation

@WillAyd could you help me understand if you have any other function/labels in mind?

[TomAugspurger]

I'd recommend moving the Series.name implementation to pandas.core.base.IndexOpsMixin, which Series and Index inherit from.

[R1j1t]

Okay, so if i understood correctly i thought something like this should work

    @property
    def name(self) -> Optional[Hashable]:
        return self.attrs.get("name", None)

    @name.setter
    def name(self, value: Optional[Hashable]) -> None:
        if not is_hashable(value):
            raise TypeError("name must be a hashable type")
        self.attrs["name"] = value

moving the above lines from pandas.core.Series.py to pandas.core.base.IndexOpsMixin should solve the problem but it only worked for series. I will try to look into your suggestion. Also my above thought didnt work.

[R1j1t]

I tried working with your suggestion of moving Series.name implementation to pandas.core.base.IndexOpsMixin but that would require a decent amount of changes (which I am not confident about).

Also, I looked at the code of pandas/core/indexes/base.py and here they validate name to be Hashable in Index class. Based on this I suggest the following:

Current version

        if name is None and hasattr(data, "name"):
            name = data.name

Proposed solution

        if name is None and hasattr(data, "name"):
            if not is_hashable(data.name):
                raise TypeError("Index.name must be a hashable type")
            name = data.name
        elif name is not None:
            if not is_hashable(name):
                raise TypeError("Index.name must be a hashable type")

If you still feel moving Series.name implementation to pandas.core.base.IndexOpsMixin would be better, then it is not straight forward to me and hence I am not proposing any solution for that. The problem I have, Series does the validation via generic.NDFrame and NDFrame is not inherited by Index. So I will have to duplicate the checking directly to IndexOpsMixin, which I wanted to avoid. Let me know if I am wrong.