Rather keep it on one level:

        if label is None:
            labels = [None]
        elif isinstance(label, str):
            labels = [label]
        elif not np.iterable(label):
            labels = [str(label)]
        else:
            labels = [str(lab) for lab in label]

Sure - I was hoping it'd have been even simpler, but str(None) returns "None", which ruined my plans.

I think you can combine this case with the one above it

Unfortunately strings are iterable, so I don't think I can...

In [1]: boo = 'who'

In [2]: import numpy as np

In [3]: np.iterable(boo)
Out[3]: True

This is exactly the code required to make int labels work.

My mistake, I thought np.iterable ignored strings. np.ndim(boo) == 0 would work, but is perhaps slower.

np.ndim looks useful. I quite readily get confused with the best way to test if something is a list, ndarray, or singleton, and our codebase uses a few incantations. I would be fine w/ changing this to something that is an agreed-upon standard way to check. This logic only gets called when a hist is added to the axes, so I don't think we need to worry overly much about performance.