TYP/CLN: Optional[Hashable] -> pandas._typing.Label

[simonjayhawkins]

No description provided.

[jbrockmendel]

LGTM

[WillAyd]

The only consideration point is that Label at least from a reader point of view was supposed to signal something that would actually be used as a Label for a row / column. Some of the uses here blur that distinction and simply leverage it because it can be used as the key for a mapping

Do you see any value in trying to maintain that distinction?

[simonjayhawkins]

can you be specific on the offending changes. where Label is used for an index name, I don't see that as being different from an row/column name.

[jbrockmendel]

@WillAyd any specific objections? i dont see any problem here

@simonjayhawkins needs rebase

[TomAugspurger]

I think this may be one case @WillAyd was thinking of. attrs needn't have any any relation with the columns of the DataFrame, so using Label may be confusing.

[WillAyd]

Yea this is a great example (thanks @TomAugspurger for following up for me!)

[jbrockmendel]

@simonjayhawkins what do you think about reverting the attrs annotation and calling this a win?

[simonjayhawkins]

will do. we have flakey ci atm but will give it a go.

(we have a fix for ci pending in #32449 but if we merge @SaturnFromTitan doesn't get credit (see #32457) or could just revert #32281 for now instead.)

[WillAyd]

lgtm feel free to merge @simonjayhawkins