REF: boilerplate for ops internal consistency

[jbrockmendel]

Progress towards #23853, with more progress available pending resolution of #27911.

[jbrockmendel]

An option for the future could make the op_name arg unnecessary (and therefore the separate steps unpack_and_defer -> _unpack_and_defer) by pinning it for the subset of methods for which it is not already correct:

@unpack_and_defer
@rename(op_name)
def generated_func(self, other): ...

instead of

@unpack_and_defer(op_name)
def generated_func(self, other): ...

[jorisvandenbossche]

For comparison ops, Index does not defer to Series

Is there a good reason for that, or just another inconsistency that ideally one time should be solved?

[jbrockmendel]

I think it has to do with wanting to make idx == ser an ndarray for the sake of ser[idx == ser] but I'm not sure

[jorisvandenbossche]

For this case, ser[idx == ser] would also work when the result is a Series

[jorisvandenbossche]

Other question for my understanding: the "unpack" in the decorator name, does that point to the unpacking of zerodim array? Or also to (future?) unpacking of Series/Index to array?

[jbrockmendel]

the "unpack" in the decorator name, does that point to the unpacking of zerodim array?

Yes. I'm open to other names if you have a strong preference.

Or also to (future?) unpacking of Series/Index to array?

I hadn't thought of that, but tentatively like it. The main thing I have in mind for extending this in the future is doing a) non-array listlike coercion and b) length-match checks since we're not very consistent about either of these.

[jorisvandenbossche]

If this is not actually unpacking our containers, I think another name would be better (I actually thought all the time those methods would be for this purpose)

Just unpack_zerodim instead of unpack can already help?

[jbrockmendel]

Just unpack_zerodim instead of unpack can already help?

Sure.

[jreback]

@jbrockmendel status of this

[jbrockmendel]

status of this

there was a request to name something differently that i need to get around to, otherwise this should be ready (havent rebased in a while, so should do that too)

[jreback]

status of this

there was a request to name something differently that i need to get around to, otherwise this should be ready (havent rebased in a while, so should do that too)

this PR lgtm. I can't find that request, what was it?

[jbrockmendel]

I can't find that request, what was it?

Addressed in a commit yesterday. It was to rename "unpack_and_defer" to "unpack_zerodim_and_defer"

[jreback]

thanks @jbrockmendel