Add dev requirements file - #4916

If we go ahead with this, I would suggest renaming the files to pip-requirements-... to make sure that they appear together alphabetically.

Good point @astrofrog - done.

No opinions have been issued thus far as to whether we should do this at all.. @mwcraig @astrofrog, @taldcroft @bsipocz, @adrn, do any of you have opinions?

FWIW, I think this is useful to have, and might make the "astropy-dev" conda install a bit easier to implement...?

Also note that this no longer requires #4915, as it turns out there's a different way to convince travis to do the right thing there. So this is now purely "information".

Not strong feelings...as noted here, I have created an astropy-dev conda environment to kick the tires on, installable with conda env create mwcraig/astropy-dev-py3.

I've opened an issue (https://github.com/conda/conda-env/issues/243) on the conda env tracker to address the fact that it seems impossible to upload an environment to an anaconda account unless you are logged in as are logged in as that user (which makes uploading to an organization hard)

No strong feelings either -- I think it could be useful, so, sure?

Just to be clear, the possible downside of this is that if the astropy-dev conda environment and RTD requirement file currently duplicate this information, so we have two different sets of requirements. Personally I think the best thing is, going forward, to call the files in this list the "official" requirements, and consider the others "derived" (even if they aren't strictly automatic). But there's still the risk of divergence.

Sure, but that will always be the case. Right now what complicates generating the conda environment a little bit are two things:

• The need to separate the dependencies (and their dependencies) into conda-installable and pip-only.
• Some conda packages are required for specific platforms but not others, so a little care is needed to generate cross-platform environments.

The first one will get easier as packages continue to be added to conda-forge; once all of the dependencies are in there we can auto-generate the environment.

Along the lines that @mwcraig mentioned, what seems slightly confusing about this approach is that many of the requirements are currently conda-installable. It isn't quite clear to me how a typical (or new) developer should actually use these requirement files. The changes to the docs in this PR just describe how to maintain the files but not how to use them.

Overall I'm definitely 👍 on capturing these requirements somehow and making it easier for developers to set up their environment.

@taldcroft @mwcraig - I agree that what I would likely do is look at the list, try to conda install <x> for each of them, and then only after I'd done that, I'd pip install -r pip-requirements-dev. That's what I basically do now anyway, but I have to remember in my head all the things that are in that file now.

Should I update the PR to include that as a recommendation? My thinking of the scope of this PR is not to update the installation instructions (for dev or otherwise), but rather to just establish the authoritative place for this information. It could be in some other format if someone has a preference, but this seems convenient if only because pip install -r <file> works at least at some level in most python environments.

This looks good to me, too.

Also the dev docs contain detailed information how to set up virtualenvs using pip only, and if anyone prefers that over conda (like me), then having the pip-requirements-dev list can be useful.

@eteq -- agree that this PR is limited in scope, so just adding this without adding anything about conda is fine.

👍

looks like it's got the A-OK, so will merge

@eteq - thanks for merging this! However, I noticed the following issue with this pull request:

• Changelog entry not present (or pull request number missing) and neither the Affects-dev nor the no-changelog-entry-needed label are set

Would it be possible to fix this? Thanks!

This is an experimental bot being written by @astrofrog - let me know if the message above is incorrect!