PyPI release checklist

[ml-evs]

Hi all, we should prepare a PyPI release once we're up to date with v0.10.0 of the spec, as the paper approaches a final draft.

This issue is intended to track packaging changes before the release, and NOT missing features/functionality of the release, which we can track at #29.

• Currently it looks like @dwinston is the maintainer on PyPI, so we might need to get that changed (not sure if Donny is still following here, so we should reach out to him)?
• Should we align PyPI major.minor version numbers with the specification? A third number can be free to track hotfixes to both the server and any changes that reflect minor fixes to the spec.
• We should decide which versions of dependencies to lock in, the important ones being Lark, FastAPI and pydantic.
• Should we switch dependency handling to pipenv? switch to pipenv? #37

[CasperWA]

* Should we align PyPI major.minor version numbers with the specification? A third number can be free to track hotfixes to both the server and any changes that reflect minor fixes to the spec.

I think this is a reasonable idea.

* We should decide which versions of dependencies to lock in, the important ones being Lark, FastAPI and pydantic.

Maybe we should even divide it out, so that one can install this package without the server, i.e., only having Lark and pydantic as the base dependencies, so that one can use the basic parts of the repo: the models and parser?
At least this makes sense for me, and would for anyone, who simply wants to use the models, but a different technology for the server implementation (although still use Python... 😃 )

[ml-evs]

Seems sensible, if we add a validator this will be another use case where all deps are not needed.

[fekad]

* Should we align PyPI major.minor version numbers with the specification? A third number can be free to track hotfixes to both the server and any changes that reflect minor fixes to the spec.

I think this is a reasonable idea.

I completely agree that the version should be related to the actual version of the specification.

* We should decide which versions of dependencies to lock in, the important ones being Lark, FastAPI and pydantic.

Maybe we should even divide it out, so that one can install this package without the server, i.e., only having Lark and pydantic as the base dependencies, so that one can use the basic parts of the repo: the models and parser?
At least this makes sense for me, and would for anyone, who simply wants to use the models, but a different technology for the server implementation (although still use Python... 😃 )

Sorry to ask, but do we really need pipy for this repository? Only a "few" people (mostly developers) will use this tool and I'm quite sure that all of them will prefer GitHub (especially when the releases are available on Github). I'm just not convinced there will be always somebody to keep pipy up-to-date just because we can.

"Pip install" can be used with GitHub if needed:
pip install git+https://github.com/Materials-Consortia/optimade-python-tools.git

[ml-evs]

Sorry to ask, but do we really need pipy for this repository? Only a "few" people (mostly developers) will use this tool and I'm quite sure that all of them will prefer GitHub (especially when the releases are available on Github). I'm just not convinced there will be always somebody to keep pipy up-to-date just because we can.

We already have one, so the question is whether we delete it or not! The question is whether this repo will ever be used for more general tools, e.g. quick command line tools that can be used to query all databases (in which case I would support a PyPI and be happy to maintain it). Otherwise, I guess I agree, as long as we make the installation instructions clear.

[ltalirz]

I would suggest we follow the practice of tagging releases from time to time and if desired, I'm happy to add a travis job that will automatically deploy new tags to PyPI (no action needed)

[dwinston]

Hi all, I still get emails on these issues, but I haven't been in the practice of reading through them as of late. I anticipated from this issue's subject that my assistance may be necessary to add owner(s)/maintainer(s) to the optimade pypi package. I need to invite by pypi.org username, so we need one or more of you to create a pypi.org account and tell me your username on this thread so I can add as an owner (who can add/remove maintainers).

[ltalirz]

thanks Donny, indeed that would be helpful - mine is leopold.talirz

[dwinston]

Okay, @ltalirz is an owner of optimade on PyPI -- I will leave it to him to add additional folks as desired because he now has all the powers.

[CasperWA]

Closing this issue, as we are on track with releasing on PyPI and will use other issues to collect information when it's necessary to release another Major or Minor version.