Visualization: dihedrals connectivity

[jngrad]

Visualizers currently represent proper dihedral bonds (A-B-C-D connectivity, aka torsion angle) as improper dihedral bonds (A-B(-C)-D connectivity, aka out-of-plane angle), causing some confusion when setting up linear polymers. This PR introduces a conditional statement to correctly treat bond connectivity in dihedral interactions.

MWE: re-use /samples/visualization_bonded.py, with box_l = 200, n_part = 20, then add

from espressomd.interactions import Dihedral
dihedral = Dihedral(bend=10, mult=1, phase=np.pi)
system.bonded_inter.add(dihedral)
for i in range(1, n_part - 3):
    system.part[i].add_bond((dihedral, system.part[i - 1].id, system.part[i + 1].id,
        system.part[i + 2].id))

right before the minimization step. Without this PR, the polymer looks like a spiderweb. With this PR, the polymer looks like beads on a string, and the magenta segments (bonds) perfectly overlap with the orange segments (dihedrals). I've tested it in OpenGL but not in MayaVi (cannot run the mayavi module in my environment because of an issue with the wx module, it has nothing to do with espresso).

@KonradBreitsprecher could you please run the MWE in both visualizers as part of the code review?

[codecov]

Codecov Report

Merging #2677 into python will not change coverage.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           python   #2677   +/-   ##
======================================
  Coverage      80%     80%           
======================================
  Files         508     508           
  Lines       27055   27055           
======================================
  Hits        21843   21843           
  Misses       5212    5212

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ca2e5f9...3c4e073. Read the comment docs.

[RudolfWeeber]

@fweik, why did you assign me?

[fweik]

Because somebody has to manage this?

[RudolfWeeber]

On Tue, Apr 09, 2019 at 05:43:08AM -0700, Florian Weik wrote: Because somebody has to manage this?

OK. I understood that differently. Manage, I can do.

[KonradBreitsprecher]

Thanks for the MWE, will test it.

[jngrad]

Update: the mayavi2 GUI works in my environment, the issue only comes when espresso makes it call the wx plugin. pip3 install wxPython --user fails because gtk+-3.0 is missing. pip3 install -f https://extras.wxpython.org/wxPython4/extras/linux/gtk3/ubuntu-18.04/ wxPython --user worked, but now there is an API error when loading the wx plugin from mayavi in espresso.

[RudolfWeeber]

@KonradBreitsprecher ?

[KonradBreitsprecher]

Tested the openGl part, no more bond 'spiderweb' there.

[jngrad]

@RudolfWeeber @KonradBreitsprecher I've removed the MayaVi fix since we cannot know if it works.

[mkuron]

Try Ubuntu 16.04 in a virtual machine. That has working mayavi packages.

[jngrad]

Tried Ubuntu 16.04 in a VM, mayavi still segfaults. Used mayavi 4.4.3 (from apt), then 4.5.0 (from pip), with traits 4.5.0, then 4.6.0.

[mkuron]

The version from Apt used to work about a year ago. Does the minimal example from #2704 (comment) also segfault?

[mkuron]

By the way, the Mac version does not segfault.

[jngrad]

I've used your MWE to test multiple versions of mayavi. Where can I find our VM image for Mac?

[RudolfWeeber]

@jngrad, could you please re-introduce the Mayavi fix and test it on someone's Mac at the ICP (@KaiSzuttor, @mkuron).
Then this PR can be finalized.