What to do about pyregion?

[cdeil]

Now that we have the effort of the regions package: https://groups.google.com/forum/#!topic/astropy-dev/QjXUl6wgM6M with some DS9 I/O functionality: http://astropy-regions.readthedocs.io/en/latest/getting_started.html#ds9 the question becomes what to do with pyregion.

There's 19 open issues and 7 open pull requests, travis-ci is broken, it looks abandoned.

There were some efforts to integrate pyregion with Astropy and improve it, especially #54 by @sargas . But those failed because there wasn't a pyregion maintainer that did code review and merge PRs and make releases.

1. One option would be to add a note to the pyregion README and top-level docs page along the lines of "pyregion is no longer under active development. Please start using and contribute to the regions package".
2. Another option would be that someone that has the commit bit for pyregion puts in the effort and make one more release of pyregion so that it can be installed properly and the PyPI page points to code and docs that exist. (see #55 #61 #71 #71 and others)

Is anyone willing to implement option 2?

I think it would be preferable, because regions is very young and not feature complete ore API compatible with pyregion and at the very least to compare regions and for tests it'll be good to have a working pyregion.

@astrofrog @leejjoon @eteq @keflavich @bsipocz @larrybradley - Thoughts?

[keflavich]

I am using pyregion on a daily basis, and while it has a lot of quirks (an API inconsistent with the rest of astropy, for example), it is powerful and necessary. So, I would favor (2). @astrofrog, might it be possible for all of the maintainers of astropy/regions to get commit priveleges on pyregion?

We might even consider making some of the astropy/regions features compatible with pyregion, e.g. a one way astropy/region->pyregion object converter.

[cdeil]

If no-one else volunteers, and I'm given the commit bit for pyregion and access on PyPI, I could implement a minimal version of option 2 early next week, i.e. get install and tests passing again and make a release. Something minimal. I'm not familiar much with the pyregion code base and don't have time to review something like #54 or judge if it's backwards-compatible enough or correct.

[keflavich]

Oof.... #54 is a lot of work. I might be qualified to review it, but I don't have the time for the foreseeable future (though maybe on a flight...)

[bsipocz]

I think both approach are fine. There are existing conda packages for the latest release, so one can use those still for some time. However if it's (2), I can help out with making the tests/infrastructure work again.

[astrofrog]

@keflavich @cdeil @bsipocz - all regions developers now have admin-level access to pyregion. In the long term we need to deprecate pyregion (once regions can do the same). It might be helpful to make a list of features that need to be implemented in regions before this can happen.

[cdeil]

@bsipocz - Thanks for fixing continuous integration for pyregion yesterday!

[cdeil]

@astrofrog - Thanks for giving us access to this repo.

So let's clean some things up and make a new release soon.

The last one is from October 2014 and as #55 remarks, points to outdated reco / docs pages: https://pypi.python.org/pypi/pyregion/1.1.4

@leejjoon - At the moment I think only you have access on PyPI for pyregion, right? Do you have time / interest to do the release?

If no, would you be willing to give the people from the new regions package access, so that one of us can make the release? See https://pypi.python.org/pypi/regions

[leejjoon]

I am willing to give others access to the package. Let me know a list of ids.

[cdeil]

@leejjoon - Thank you!

My suggestion would have been to use the same list as https://pypi.python.org/pypi/regions

Package Index Owner: keflavich, lbradley, robitaille, Christoph.Deil

Or, if you prefer, just add me for now, and I'll make a 1.2 release in the coming days and then add others later as needed.

[leejjoon]

This should be done now.

[cdeil]

I'm closing this discussion issue now. Thanks to everyone here for your input!

In the past days we've agreed on and largely implemented a solution.

For pyregion the continuous integration setup has been fixed, there's a link to the new regions package in place, and the pyregion v1.2 release is coming tomorrow.