Future of the mayaviLive visualizer in ESPResSo

[Jean-Noël Grad]

Dear ESPResSo users,

The ESPResSo python interface currently supports two visualizers: one 
based on OpenGL (openGLLive) and one based on Mayavi (mayaviLive). The 
OpenGL version is actively developed and supports almost all features of 
ESPResSo. On the contrary, the Mayavi visualizer is no longer developed 
and doesn't implement many features. For example, it doesn't support 
drawing shape-based constraints and LB fluids, and it is unclear whether 
it properly supports collision detection and bond breakage.

In our experience, the dependencies of the Mayavi visualizer are also 
more difficult to install than the dependencies of the OpenGL 
visualizer. Currently, the list of dependencies for Mayavi cannot be 
satisfied on the workstations used at the home institution where 
ESPResSo is actively developed. This means that the ESPResSo team cannot 
verify whether mayaviLive works. In particular, the occasional changes 
made to the mayaviLive code over the last two years to reflect changes 
in the Particle data structure were made without verifying if the 
visualizer was still functional.

Maintaining multiple visualizers can benefit end users in the following 
ways:
1. better platform support, e.g. if the dependency tree of visualizer X 
cannot be satisfied on a specific workstation, visualizer Y can be used 
as a fallback
2. more diverse workflows: mayaviLive can be connected to other 
components of the Mayavi python ecosystem to customize visualization

Regarding point 1, we know the community is already using alternative 
visualization tools to inspect ESPResSo trajectories, namely VMD, 
ParaView and Blender, either because it is more convenient to use these 
external tools, or because they provide analysis methods to post-process 
the data, or because they allow producing professional-looking 
ray-traced movies. Therefore, if openGLLive cannot be used in a 
particular software environment, the community already has access to 
multiple fallback solutions. Regarding point 2, we are currently unaware 
of visualization workflows that couple mayaviLive to external python 
packages.

Resolving the mayaviLive dependency tree would require us to recompile 
the C++ GUI library wxPython from sources and resolve several 
compatibility issues between the Mayavi python package and its 
dependencies. This developer time could be better invested in improving 
openGLLive and other ESPResSo features. In addition, mayaviLive depends 
on C++ global variables, making it more difficult for us to remove the 
global state, which is a necessary step for our long-term goal of 
allowing multiple ESPResSo system instances in the same simulation 
script.

We currently do not know if mayaviLive is used in the ESPResSo 
community. To help us make an informed decision about the future of this 
visualizer, we kindly ask you to reply to this email if you are using 
mayaviLive, preferably until Wednesday, May 11th 2022.

Best regards,
Jean-Noël Grad

Institute for Computational Physics
University of Stuttgart
Allmandring 3
70569 Stuttgart
Germany
Phone: +49(0)711/685-67715
Email: jgrad@icp.uni-stuttgart.de
https://www2.icp.uni-stuttgart.de/~icp/Jean-No%C3%ABl_Grad