ratt-ru/CubiCal

Why not in pypi?

gijzelaerr opened this issue · 10 comments

gijzelaerr commented

Why is cubical not in pypi? makes it easier to install.

I've taken the liberty to upload the latest version, but that fails:

HTTPError: 400 Bad Request from https://upload.pypi.org/legacy/
Invalid value for requires_dist. Error: Can't have direct dependency: "montblanc @ git+https://github.com/ska-sa/montblanc.git@0.6.1 ; extra == 'lsm-support'"

Montblanc is actually on pypi, but not version 0.6.1. Any reason why 0.6.1 specifically is required, or can we just upgrade.

o-smirnov commented

Tender triggering tremours of tensorflow here, or maybe I've had too much coffee.

gijzelaerr commented

that caused by an optional dependency on Montblanc, but that doesn't mean it can't be uploaded to pypi.

made a little pr:

#446

and uploaded it also:

https://pypi.org/project/cubical/

JSKenyon commented

There used to be more reasons why we weren't on PyPI - Montblanc is the final sticking point. It is technically end of life now, but if @sjperkins is feeling charitable, maybe he can help us out.

JSKenyon commented

@gijzelaerr Could you please make me a maintainer of CubiCal on PyPI? Unless you are ok with me prodding you to do version uploads.

I guess we could automate it to happen on release, though I have never done that.

bennahugo commented

That is quite trivial to do. All you need is to store the secret token in github secrets and add a github action

e.g.
https://github.com/ska-sa/meqtrees-cattery/blob/master/.github/workflows/python-publish.yml

gijzelaerr commented

what is your pypi username

JSKenyon commented

JSKenyon

gijzelaerr commented
JSKenyon commented

Thanks! Accepted.

bennahugo commented

I believe this is resolved since