[WIP] Update highlighted points when time changes

[andy-sweet]

Closes #53
Closes #80

This updates the highlighted points when the time point changes. It also distinguishes between the number of loaded tracks and highlighted points at the current time point, presenting those to the user.

It does this by storing the point IDs in the tracks with the start and end time (which are all readily available), then using that to convert point IDs to time-specific indices.

This is probably the smallest change that gives us the behavior described in #53. However, this does against some of the ideas in #53 and #68. I agree that mixing ID conversion logic and geometry information is undesirable, so very open to other approaches here.

The approach here also has performance issues when a large number of tracks have been loaded because the conversion takes a little time.

zebrahub-update-selected-points.mp4

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
points-web-viewer	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jun 5, 2024 8:40pm

[andy-sweet]

I don't think this reviewable/mergable yet. The behavior might be good enough, but there are some performance issues. I'd also want to wait to merge #95 before marking this ready for review.

[aganders3]

This is probably the smallest change that gives us the behavior described in #53. However, this does against some of the ideas in #53 and #68. I agree that mixing ID conversion logic and geometry information is undesirable, so very open to other approaches here.

My idea here would be to instead store a(nother) map on PointCanvas that maps curTime -> pointIndices (number -> Set<number>). This could be filled in during canvas.addTrack. Memory usage should be similar but it should be a bit faster to update, as there'd be no calculations and no loop over tracks when changing the timepoint.

I'm also trying to think about doing it in the shader (create a new PointsMaterial to support highlights), but I think it would be more complicated, less general-purpose, and it might not even be faster in the end.

(apologies - I know you said this is not ready for review)