Release 1.0.1

[orbeckst]

A few things are accumulating for a patch-level release for 1.0, related to

• installation
• performance regressions

The point is to make 1.0.x a stable platform for anyone who needs legacy MDAnalysis for the foreseeable feature. 1.0.x should not be worse than 0.20.1.

[orbeckst]

One potentially controversial statement in Challenge to scientists: does your ten-year-old code still run? Nature 2020 doi:10.1038/d41586-020-02462-7

Newer languages’ rapidly evolving APIs and reliance on third-party libraries make them vulnerable to breaking. In that sense, the sunsetting of Python 2.7 at the start of this year represents an opportunity for scientists, Rougier and Hinsen note. Python 2.7 puts “at our disposal an advanced programming language that is guaranteed not to evolve anymore”, Rougier writes (Rougier, N. P. ReScience C https://doi.org/10.5281/zenodo.3886628).

provides some additional motivation for making sure that we have a stable Python 2.7-capable release.

[orbeckst]

The backports from PR #2798 are now included.

The remaining major issues is to either disable capped-distance with nsgrid #2930 or fix it.

[jbarnoud]

The fix for #2984 could be backported.

[IAlibay]

So after some off-github discussions with @lilyminium and @richardjgowers, we were wondering how everyone would feel about pushing for a rapid 1.0.1 release where we disable nsgrid completely and then have aim for a 1.0.2 release with @richardjgowers' fix?

If the mailing list is any indication, quite a lot of people are facing issues with things that are now fixed in master, on top of that we've got this nsgrid bug that users probably aren't aware of. Personally I think it would be better for everyone to have some type of patch fix out there earlier than later?

pinging @MDAnalysis/coredevs here.

[lilyminium]

+1 for releasing an interim 1.0.1 as strictly better than keeping 1.0.0 until the fix is in (e.g. all the pytest failures)

[orbeckst]

Yes, I agree – disable nsgrid #2930 for 1.0.1.

I'd also include a backport fix for #2984 and then do a 1.0.1. We can really do a patch release in rapid succession. Doubt that anyone minds getting code fixed.

[lilyminium]

And #2427!

[orbeckst]

I'll backport PR #2427.

[orbeckst]

Outstanding backports/fixes

• disable nsgrid neighbor search #2930
• backport PR Always transplant a stub for transplanted methods #2427
• backport AtomGroup.center fails for compounds+wrapping #2984
• backport PR Add dev and stable versions once #3009 (fix docs version selector has multiple repeated 'dev' entries #3006 )

[orbeckst]

The deprecation of hbond.waterbridge analysis in PR #2913 could go in 1.0.1 if it gets turned around quickly. Otherwise we make a 1.0.2 soon-ish.

[orbeckst]

Unfortunately, I broke Travis for master (see PR #3011) so that PR also has to go through before we can wrap 1.0.1.

[IAlibay]

Just FYI unless there's any objections, I'm going to declare a code freeze on 1.0.1 (excluding release related CI/maintenance stuff).

There's a lot of things that still need deprecating/backporting (e.g. #2913, #2857), and then the nsgrid fix, so we'll need to cut a 1.0.2 release for them.

I still have hopes we can release this week, assuming we don't encounter more upstream issues...

[IAlibay]

Release 1.0.1 has been completed -- closing