evaluate mol* for replacement

toloudis commented 5 months ago

What needs to happen?

look at mol* code and determine what it would take to actually replace the simularium-viewer canvas with their canvas and drawing api.

how do we glue our data structures to their renderers inputs

Why should we do this?

To eval the size of effort

When does this need to get done?

within the next couple of simularium dev mtgs

toloudis commented 4 months ago

After some eval, I think it might be a reasonably small few days of focused time project to try to wedge the mol* canvas into the viewer. as a basic starting point for eval / comparison. we won't automatically get all of our same features back. But there seems to be a clear path for us to be able to render the same meshes and pdb sphere collections that we show now. TBD:

instancing support and perf
AO settings for scenes with widely varying spatial scales
re-enable outlining
instanced curves/fibers
agent selection/picking
plug in to their LOD system
plug in to their transparency system Big questions: performance of scene updates across timesteps. performance of large numbers of instanced geometries such as whole pdb molecules.

As previously discussed, we can also go the other direction and ask: what exactly do we want from their codebase for the simularium viewer?

toloudis commented 2 days ago

There were some blockers with this, with respect to using the mol codebase to get anything to render. The mol devs addressed this initial blocker but we never revisited.

Current considerations (while we move forward with simularium viewer dev tasks):

mol* is clearly way ahead of our viewer on loading/parsing/displaying pdb molecular geometries
mol* is clearly behind our tech on displaying volume data
mol* is not as well optimized for time-series full-scene updates
mol* is a more general rendering system and is probably ahead of ours on core rendering features (some comparison tbd)
the mol* renderer code is a bit complex (and not well documented for integration tasks like this) -- and is a nontrivial time investment to prototype with simularium data

toloudis commented 1 day ago

For reference, prototype code branch is here: https://github.com/simularium/simularium-viewer/tree/feature/molstar-canvas

simularium / simularium-viewer