Release 1.1.0

[IAlibay]

Hi all,

Having recently completed 1.0.1, we would like to quickly deliver 1.1.0 ASAP.

This is because:

1. Windows does not currently work with 1.0.1
2. There are lots of very useful features sitting in 2.0 which cannot be used (RDKIT, MSD, new helanal, etc...)
3. Maintaining the 1.x branch is becoming unmanageable - we are now at a point where even py3.6 is getting dropped by upstream packages.

In order to speed things up, we would like to propose a 1 week deadline for code fixes. That is to say, unless there is a good reason not to, after the 21st of March 2021, we will put a code freeze on the current master branch and move towards making a release.

After discussions with @richardjgowers and @lilyminium, it is our view that anything that requires longer than that should probably just go into 2.0. Please do speak up if you disagree with this.

As it stands, this is the list of things I believe were earmarked for a final 1.x release:

• TPR indexing TPRParser should not index from 0 #2364 (PR Add tpr_resid_from_one keyword to TPRParser #3153)
• Small boxes in capped_distance capped distance fails for very small (<0.1A) boxes #2670
• Warning about upcoming changes to the NCDFWriter's default behaviour (w.r.t. velocity scale_units) Improving the NCDFWriter behaviour (units + scale_factors) #2327 PR Issue 2327 - Warn of upcoming scale_factor changes #3185
• Temp/B factors (I don't think we need warnings for this, but @lilyminium was saying otherwise?) (PR Added deprecation warning when trying to add_TopAttr bfactors or tempfactors #3161)
• Dealing with hbond_autocorrel (@p-j-smith based on HydrogenBondAutoCorrel uses deprecated features of numpy #2987 (comment) are we ok to just deprecate or should we just move it?) - @richardjgowers to fix and maybe move instead.
• Warning for changes to the way we get bonds/angles/dihedrals from atomgroups (u.select_atoms(...).bonds should return only bonds with atoms *inside* of the selection #2821 and Dihedrals and selections #1264) (PR Added get_connections #3160)
• Fixing Windows builds (abspaths slipping into SOURCES and breaking windows conda build #3129)
• Parmed ts deprecation Remove support for ts as an input to the ParmedConverter #3031
• Warning for PDB segid changes Write out chainID to PDB instead of segID #3157

Is there anything else we need to do? Is this reasonable for everyone?

pinging @MDAnalysis/coredevs

[lilyminium]

re: temp/bfactors -- we could warn for the extremely rare case that someone has used add_TopologyAttr so that both factors are in the same universe, but contain different values, possibly as tags. I'll see if I can make a warning for that case only (that both exist).

[orbeckst]

1.1.0 is then really the final MDA 1.x release (short of errors in the package build which might require a patch release).

We'll then stop any support for Python 2.7 & 3.6 (yikes... I am unhappy about how fragile the Python ecosystem is... but that's a different rant) and no more bug fixes for 1.x.

(EDIT: The above just restates what https://www.mdanalysis.org/2019/11/06/roadmap/ said: development on 1.x will cease with the release of 2.0)

We should publish a requirement file with last known good versions of all deps that make 1.1.0 work.

[IAlibay]

With no surprise to anyone, looks like we'll lapse our deadline, but most PRs (except from any NSGrid changes) are open. So unless there's any objections, let's just say it's a code freeze on any new non-critical PRs. If @MDAnalysis/coredevs can concentrate on reviewing what's open we can move with a release as soon as possible.

[IAlibay]

Ah seem to have forgotten about hbond_autocorrel. @richardjgowers how do you want to deal with it?

[tylerjereddy]

I am unhappy about how fragile the Python ecosystem is

Yeah, the release cadence is 1 major Python version per year now I think.

We should publish a requirement file with last known good versions of all deps that make 1.1.0 work.

We're working on doing something similar upstream--pinning build- and run-time deps in releases for long-term repeatable builds. For example: scipy/scipy#12862

[IAlibay]

What's left to do here?

@richardjgowers - is there anything that needs to be done for NSGrid & hbond autocorrel?

[IAlibay]

Release 1.1.1 is now available (we had to skip over 1.1.0 due to a pypi upload issue) 🎉