blog post for 0.19.0

[orbeckst]

The 0.19.0 release (MDAnalysis/mdanalysis#1971) is imminent (PR MDAnalysis/mdanalysis#2095) so we need a blog post that summarizes the major things.

• Windows conda packages (Python 3 only)
• contributions from GSoC (we want the separate posts Blog post for on-the-fly transformations #87 and GSoC Blog Post : Improved Distance search methods #90 up as soon as possible, too, so that we can link to them; I would prefer them to come before the release post)
  • distances and PBC handling
    – link to GSoC Blog Post : Improved Distance search methods #90
    - periodic keyword for select_atoms()
  • on-the-fly transformations – link to Blog post for on-the-fly transformations #87
• contributions from NSF REU (dihedral and Ramachandran/Janin analysis – can use something from Henry's bio page)
• various performance improvements (see CHANGELOG)
• ChainReader can now smartly concatenate trajectories and drop duplicated frames
• sliced trajectory objects
• improvements/additions in analysis
  • density (more flexibility in choosing the analyzed volume)
  • radial distribution function (InterRDF_s)
• formats
  • added hybrid36 format for PDB-like files
  • SETTLE constraints in Gromacs TPR files are now read as bonds (Gromacs water defined with SETTLES appear as disconnected when read from TPR mdanalysis#1949)
  • MemoryReader is now feature-complete: supports box size and the new transformations; various inconsistencies fixed. Basic idea: any changes to the MemoryReader's data (positions, velocities, forces, dimensions) are permanent.
• deprecations
  • removed quiet=True; use verbose=False instead
  • analysis: use run(start, stop, step)
  • save data using your own tools instead of .save() methods: easier to maintain because formats change (e.g. pickle) and we think the user knows best how they want to keep their data
• deprecated "flags" (probably no-one used them... remove flag registry mdanalysis#782 for details)
• highlight new contributors

[richardjgowers]

Yeah it would be good to get the two gsoc posts finally finished, I'm behind on that

[orbeckst]

In discussion with @davidercruz on PR #87 it became clear that some of the crucial features for on-the-fly transformations are not in 0.19.0 yet but are still PRs

• wrap/unwrap PR Added unwrap/wrap transformations mdanalysis#2038
• fitting transformations PR Added fitting transformations mdanalysis#1991
• centering PR plane and axis centering transformations mdanalysis#1973

@davidercruz rightly pointed out #87 (comment) that announcing OTFT without these features makes it much less exciting (and robs it of impact).

My suggestion for the 0.19.0 blog post is #87 (comment) to only briefly mention the transformations along the lines that we outline the idea, state that the foundations are implemented, it's experimental, and cool things are coming.

The OTFT blog post #86 (PR #87) will then be published after the 0.19.0 release announcement and we try to get the OTFT PRs done for a 0.20.0 (or 1.0.0?) release. (We cannot put them in a 0.19.1 because they are new features and those cannot go into a patch release under semantic versioning.)

What are your opinions, especially @davidercruz @jbarnoud @richardjgowers ?

[jbarnoud]

Sorry, These PRs are hanging by my fault. The wrap/unwrap is basically ready. I would need to run it on a couple of systems to make sure I did not miss anything obvious but I can do that latter today. The two other PRs need a bit more testing on my part as they are doing less obvious things.

[davidercruz]

The PRs are really close to completion, so, unless some nasty bugs are found, we can try to push OTFT for the 0.19.0 release. @jbarnoud what do you think?

[richardjgowers]

@davidercruz 0.19 is out already.

We can wait and make sure all these PRs are 100% finished and make 0.20.0 the big transformations release. It won't be another 6 months before a release this time, I'm working on making the process easier

[orbeckst]

@richardjgowers we should get this blog post updated for 0.19.x (Windows!) and then do an announcement on twitter, too.