AnimationAction: action.enabled doesn’t change action._effectiveWeight

[jostschmithals]

There seems to be a logical inconsistency in AnimationMixer/AnimationAction:

From examining the comments and the code structure it becomes clear, that the action.paused property and the action.enabled property should work analogous, one concerning timescale/warping and the other concerning weight/fading.

Setting action.paused to true changes the value of action._effectiveTimeScale temporarily to 0 and leaves action.timeScale untouched (important for being able to restore this value when unpausing).

This is correct.

Setting action.enabled to false should actually do the same for the weight: leaving action.weight untouched, but changing action._effectiveWeight to 0.

But this doesn’t work.

Setting action.enabled to false has no effect on action._effectiveWeight. This is reflected for example in the dat-gui of webgl_animation_skinning_morph.html (for reproducing please deselect “enabled” in the “Clip’Action’” section and watch the “eff weight” slider).

The reason:

On the one hand action._updateWeight is implemented correctly (analogous to action._updateTimeScale), so that action._effectiveWeight actually should be set to 0 on setting enabled to false. - But: action._updateWeight will never be executed in this (action.enabled = false) case, since it is only called by action._update, and this is only called by mixer._update, if action.enabled is true.

Nevertheless the knight model in the example behaves correctly at the first glance (as if the effective weight would be set to 0 on disabling). But this happens kind of accidentally, due to the fact that the further processing of the weight (the correct property binding) is interrupted in this case. I think that is not how things should work under the hood?

How to fix:

Probably simply remove in mixer._update the condition: if (action.enabled) ? – Seems to work without negative side effects.

---

By the way:
Two mistakes in the comments should be fixed, too:

1. In the constructor of AnimationAction:
   “this.enabled = true; // true -> zero effective weight” (must be “1” instead of “zero”)
2. In setEffectiveTimeScale():
   “// set the weight stopping any scheduled warping” (must be “set the time scale …”)

[mrdoob]

/ping @tschw @takahirox

[takahirox]

It seems like a bug because
_updateWeight() takes care of enabled

three.js/src/animation/AnimationAction.js

Lines 382 to 408 in 08c40e2

	if ( this.enabled ) {
	weight = this.weight;
	var interpolant = this._weightInterpolant;
	if ( interpolant !== null ) {
	var interpolantValue = interpolant.evaluate( time )[ 0 ];
	weight *= interpolantValue;
	if ( time > interpolant.parameterPositions[ 1 ] ) {
	this.stopFading();
	if ( interpolantValue === 0 ) {
	// faded out, disable
	this.enabled = false;
	}
	}
	}
	}

but it won't be called if enabled is false.

I'll try to make PR to fix it...

[jostschmithals]

Just to support your opinion (assuming that this really is a bug):

Line 146 states: … .enabled = false yields an effective weight of zero …

three.js/src/animation/AnimationAction.js

Line 146 in 08c40e2

// although .enabled = false yields an effective weight of zero, this

And lines 152/153, read together, are implying the same:

three.js/src/animation/AnimationAction.js

Lines 152 to 153 in 08c40e2

	// note: same logic as when updated at runtime
	this._effectiveWeight = this.enabled ? weight : 0;