cf-python
cf-python copied to clipboard
NCAS-CMS
Documentation request: clarity as to which versions are currently available where
The current version is 3.4.0. I have 3.0.1 in a (relatively recent) conda install via the ncas channel. It would be good to have clarity about which versions are available where.
I agree. PyPi is always up to date, but conda sometime lags behind. We'll have a re-thnk about our workflow for releases.
You can always pip install into your conda installation (which is what I usually do!)
It would be good to have clarity about which versions are available where.
Hi @bnlawrence, I also agree this would be useful, & whilst I was putting up some reference information I realised there is quite a simple way to do it, namely via 'shields' which are little badges you have probably seen around on GtiHub or elsewhere. They query relevant APIs so they can provide up-to-date information, notably for their versioning support they query PyPI and conda so can automatically pick up on the versions hosted in both locations (even under our ncas channel which is quite neat).
So, whilst adding in the shields for the references (which was the task I was originally doing) I added shields with installation information & they do display the latest versions available on PyPI & conda. You can find that now at the top of the README & it will always be up-to-date, so I suggest you bookmark it here at https://github.com/NCAS-CMS/cf-python#cf-python and check it whenever required (I've added the equivalent for cfdm too so that can also be examined).
Does that seem like a sufficient solution to you? If so I can close this (though probably I should also add a note in the documentation itself linking to the top of the README to state that the versions available are & will always be provided accurately there, so it is can become open knowledge).
PyPi is always up to date, but conda sometime lags behind. We'll have a re-thnk about our workflow for releases.
We should still address this, of course! But this is more about talking to Andy to agree how we can help to manage the ncas conda channel.
Thanks, @sadielbartholomew - the tags in the README are just what we needed.
Glad they work for everyone. @davidhassell if you want to edit those little badges I put up in any way (add more, remove some, change the look & feel) feel free, the library used is shields & the site helps you set them up. I changed the colours because by default they come out differently for the various shields & are quite bright & garish IMO.
I noticed during this that cfdm does not have a conda sub-section in the installation docs. Is there a reason for that, or shall we add one?
Thanks. The reason for the lack of a conda section in the cfdm docs is pertinent, as I didn't put it in because I didn't want to advertise an out-of-date version!
The cfdm situation is not quite the same as cf-python, however, because the installation dependencies are very light for the former, and all available via pip. For cf-python, though, we need the not-in-PyPi libraries udunits2 and ESMF, so there is an advantage in using conda ... if only it were always up to date!
Just to be clear, what is the motivation for re-opening? Is it to ensure the conda versions are up-to-date? If so we could amend the title of the Issue so it reflects that, to help us keep track.
Oh, sorry - I hadn't really thought it through. We either close it and accept the status quo, but the version in conda is not recorded in the installation page. So I think your idea of renaming the issue to reflect this is a good, to see if we can come up with something better in consultation with @ajheaps.
We either close it and accept the status quo, but the version in conda is not recorded in the installation page
No worries & yes that is true, so apologies it was probably a premature close. I was going to just add a 'note' to the installation page to say users can find the versions on the README with a link there but if you prefer some other solution for the docs that is fine. But yes I think the key now is trying to keep the conda versions up-to-date, at least more so than before.