PyPI package is old about aipy HOT 16 CLOSED

3.0.0rc2 is actually available on PyPI as a prerelease:

https://pypi.org/project/aipy/3.0.0rc2/

So I believe that with the right flags you should be able to get pip to install it. However, that prerelease has been out for a long time at this point ... we should probably just make it official and call it 3.0.0.

from aipy.

Ah, OK. Interesting -- so PyPI won't try to install an rc version unless forced to, while conda will. I'll change the flags!

from aipy.

After reading some backlog conversation on the herasoftware slack channel, looks like this is the same as the issue I just posted (#56). Who has permission to make 3.0.0 official? @pkgw @AaronParsons ?

from aipy.

@adampbeardsley I believe that I have the technical permissions, but I'd want to get the OK from @AaronParsons first.

from aipy.

OK, so I just checked the actual setup.py at https://pypi.org/project/aipy/3.0.0rc2/, and it does not have the conditional that chooses between py2 and py3 -- it is hard-coded to depend on 'astropy>=1.0, <2.0', which will definitely fail for py3. That makes bumping the PyPI version very urgent.

from aipy.

The Astropy version constraint was given by the person who contributed the port of AIPY to Python 3 — he said that it didn't work with Astropy >= 2. Unfortunately I don't recall what the issue was. Given where things are, I think the pretty clear path forward is to relax that requirement and figure out fixes for any breakages related to newer Astropys if and when they are uncovered.

(That being said ... I have trouble imagining that the incompatibility was the sort of thing that would lead to subtly wrong science results rather than code crashes, but maybe it is!)

from aipy.

Hmmm. That is strange. Astropy >=3 certainly doesn't seem to break aipy (in any use case I've found) on py3.

from aipy.

@pkgw @adampbeardsley Just bugged @AaronParsons about this. He said it's fine. Also he's in favor of adding someone new to have the permissions for aipy on PyPI. @plaplant was nominated on today's pyuvdata telecon and both he and @AaronParsons are fine with that. @pkgw can you extend permissions to @plaplant?

from aipy.

@jsdillon Looks like I have permissions to make new releases but not to manage the list of project maintainers — @AaronParsons will have to do that personally.

from aipy.

The Astropy version constraint was given by the person who contributed the port of AIPY to Python 3 — he said that it didn't work with Astropy >= 2. Unfortunately I don't recall what the issue was.

This may be a little late but the whole 'astropy>=1.0, <2.0' thing is/was about maintaining compatibility with Python2 (astropy dropped support for it starting with version 2.0). If Python2 support doesn't matter then the version of astropy is flexible.

from aipy.

@jaycedowell Thanks, that's helpful. In that case we should be totally OK to relax that constraint.

from aipy.

@pkgw I think this is still open? Are you able to bump the version on PyPI?

from aipy.

@adampbeardsley I do have permissions to upload a new version. To make this sustainable we need an intervention from @AaronParsons on PyPI to give @plaplant ownership and myself privileges on the package, though.

from aipy.

@pkgw can you upload a new version in this one instance? @AaronParsons is away right now, and the astropy version constraints are stalling our builds for dependencies. We should have things in a more sustainable configuration sometime next week.

from aipy.

OK, I've tidied up the current repo and published 3.0.1.

from aipy.

Thanks so much @pkgw!

from aipy.