world-aware adding of particles to fluid system about salva

[sebcrozet]

Hi! Testing if a particle exists at a specific location isn't implemented yet on Salva, so for now testing ball-ball proximity is the only option unfortunately.

Testing if there is a collider in a specific location is possible if you are using nphysics because you can use geometrical_world.interferences_with_point. And if you are using ncollide directly, you can do collision_world.interferences_with_point.

[rezural]

OK, Thanks.

Any ideas on data structures for efficient search through the particle space? Would this structure need to be updated for every particle movement? Would it be too prohibitive?

Cheers

[sebcrozet]

@rezural The ideal data structure would be a grid. Actually, this is what the LiquidWorld uses and maintains internally. So the ideally we would want to let the LiquidWorld tell us what particles lie in a given region of space.

Alternatively, you can maintain your own HGrid by clearing and re-adding points after each update and use HGrid::cells_intersecting_aabb to find all the particles in a given area (defined as the AABB of the particle you intend to add).

[rezural]

Right,

Something like?

liquid_world.particles_within(region: Shape, isometry) -> Vec<Point3<f32>>
liquid_world.particles_within_aabb(region: Shape, isometry) -> HGrid<..> // for efficient lookup 

I can probably add this in there of you'd like to see it.

[sebcrozet]

The following should be enough:

liquid_world.particles_intersecting_aabb(&self, aabb: &AABB<f32>) -> Vec<Point<f32>>;

In the end we would probably need to return some identifier that allows the user to retrieve the fluid containing said particle, but since fluid/boundary addressing will change soon, I think returning Vec<Point<f32>> is a good start and will cover your use-case.

I can probably add this in there of you'd like to see it.

Yes, that would be great. Ideally, the PR should target the rapier branch.

[rezural]

So the rapier branch doesnt build against rapier (with rapier, ncollide, nalgebra cloned locally).

You mentioned a salva branch on the rapier repository? I cant find that on the dimforge repo.

Otherwise I will make a pr against salva master, hopefully within a few days.

[sebcrozet]

You mentioned a salva branch on the rapier repository? I cant find that on the dimforge repo.

That's right, I forgot to push it. Today I pushed it and merged it to the master branch of Rapier.

In addition I merged #11 which drops nphysics to use Rapier instead. So the master branch of salva now compiles and has integration with Rapier (using the github version of Rapier until it is released at the end of this month).

[rezural]

OK, thanks for that!

I will look into getting this working with the latest master.

[rezural]

I'm tempted to add a particles_intersecting_aabb method to the Fluid struct, and just collect these together in the liquid world, which will probably make this easier to work in with the addressing changes.. Thoughts?

[sebcrozet]

This will be implemented by #13