Plans to publish `signatures` to pypi?

[tlambert03]

Hi 👋

love what's happening here.

I'm playing around with stubbing out protocols for common image analysis operations (e.g. scikit-image), and would love to build on what already exists in API_specification/signatures. I'm wondering if you have any plans to publish/distribute the stubs in that folder to PyPI?

[honno]

I'd be interested in this for the test suite too, where we utilise signatures already. Might help us with https://github.com/data-apis/array-api-tests/issues/107.

I imagine this could be a completely automated process from this repo (i.e. no separate repo), with calender versioned releases. Have to be mindful of #398—possibly the top-level package could be namespaced with each spec release, e.g. xp_signatures.v1. Probably worth waiting until we get a first spec release generally.

[rgommers]

Probably worth waiting until we get a first spec release generally.

We now have a final release. This isn't too much work to do. Should we go ahead with this?

[honno]

I'd propose we first work on setting up a workflow where the signatures folder is uploaded as-is, available at say the <pkg_name>.draft namespace. Then we could get onto uploading versioned namespaces, which might be a bit annoying but still quite do-able.

[kgryte]

@honno Is this something that we can move forward?

[honno]

@honno Is this something that we can move forward?

I think we'd really want to get in https://github.com/data-apis/array-api/pull/589 first (which LGTM but we'd want someone else to give the go-ahead, then we can ping nstarman to see if they can update/rebase the PR (or if not available I can have a go).

Then we could publish basically as-is, or I could update https://github.com/data-apis/array-api/pull/472 with said changes assuming folks are happy with the direction I was going for... I'll have a think over but again #589 is blocking anywho IMO.