Timer: Added FixedTimer

[mrdoob]

Related issue: #17912

Description

Moved the _fixedDelta and _useFixedDelta functionality into its own FixedTimer class and removed the enableFixedDelta() and disableFixedDelta() methods.

I think this approach is cleaner... but I'm not sure how FixedTimer is supposed to be used.

@Mugen87 I guess it's supposed to be used for debugging? #17912 (comment)

[CodyJasonBennett]

Very nice for physics or delta-based simulation, also see https://gafferongames.com/post/fix_your_timestep.

[mrdoob]

Hmm...

In the context of physics in the browser, isn't this better?

const fps = 60;

setInterval( step, 1000 / fps );

function step() {

    world.stepSimulation( 1000 / fps, 10 );
    ...

}

https://github.com/mrdoob/three.js/blob/dev/examples/jsm/physics/AmmoPhysics.js#L265
https://github.com/mrdoob/three.js/blob/dev/examples/jsm/physics/RapierPhysics.js#L209

Otherwise, FixedTimer would need to have more logic inside if we want to ensure 60fps in any device.

[Mugen87]

The idea of fixed time deltas comes from Unity: https://docs.unity3d.com/ScriptReference/Time.html. They describe their property as follows:

The interval in seconds at which physics and other fixed frame rate updates (like MonoBehaviour's FixedUpdate) are performed.

I have found it useful by myself if I can configure a fixed time delta for testing purposes.

[Mugen87]

Does the setInterval() approach compute exact time deltas? I'm asking since I don't know how precise setInterval() actually is.

[mrdoob]

Using setInterval() we can update a function at 60fps even if the device runs at 120fps.

Then we compute the deltas ourselves:
https://github.com/mrdoob/three.js/blob/dev/examples/jsm/physics/AmmoPhysics.js#L209

I can imagine new FixedTimer() being useful for debugging, but I'm just not sure how it's supposed to be used for physics.

[CodyJasonBennett]

I had misunderstood how this class worked. Physics wants to subdivide the real delta time as per the reference I linked, which is not what this class does.

[Mugen87]

@mrdoob Indeed, we can't use it for physics but FixedTimer still has its value for testing purposes. Do you think this use case justifies to keep FixedTimer?

The refactoring itself definitely makes sense and the PR in its current state looks good to me (just need to clean up the docs).

[mrdoob]

I think we can keep FixedTimer yeah.

I'll remove the references to fixedDelta in the docs.