Added "controllermove" event to WebXRController

[arodic]

This is a requirement to make #20720 work in WebXR.

The problem it solves:

Right now WebXRControllers emit various events such as select, selectstart, selectend, squeeze, squeezestart, squeezeend, pinchend, pinchstart, connected and disconnected. But no event is fired when controllers move. The only way to get controller movement is to continuously pull transformation from rAF.

I would like to be able to use controllers in a similar fashion as other input devices (mouse, touch, pen, keyboard...) strictly via event listeners. I think this PR will enable that at least for simple click (select) and drag gesture.

[mrdoob]

Closed?

[arodic]

Closed?

I merged into it to #20720 source branch. Do you prefer to keep it in a separate PR?

[mrdoob]

There's another PR enhancing WebXRManager: #20696
The more complex the PR = the more conflicts there will be = the longer will take to be merged.

[arodic]

Ok, closing #20949 in favor of this one.

[mrdoob]

Realistically, don't vr controllers move all the time?

[arodic]

Realistically, don't vr controllers move all the time?

Yes @mrdoob, while they are active in the session they do move on every frame.

In the context of XR support for TransformControls, I am trying to get XR controller input events, exactly like pointer and keyboard input. We already have events selectstart and selectend which are equivalent to pointerdown and pointerup from UX perspective. However, there is no equivalent to 'pointermove`. The problem is that without that event, TransformControls would need to do one of the following:

1. Run internal XR-rAF all the time to get controller movement.
2. Start internal XR-rAF on selectstart to get movement values and stop it on selectend.
3. Require application to call updateXRControllers() externally from the application rAF.

Options 1 and 2 are not ideal because suddenly it requires internal animation loop to get user input. Option 3 is the other side of the same coin, it requires application to continuously update() controls which was not necessary for pointers.

I'm not that familiar with WebXR APIs but I find the design of the event system somewhat incomplete. We have some events for two primary (special) controller buttons - selectend, selectstart, squeeze, squeezeend and squeezestart - which nicely maps to familiar pointer/keyboard down/up events but it is missing the key *move event. At the same time, all other buttons are mapped to the GamePad API which afaik does not have button events at all and requires continuous pulling of buttons states.

Ideally, I would love to be able to use XR Controllers and their GamePads exactly like mouse and keyboard, by attaching event listeners. Would that ever be possible? Was that discussed for the WebXR spec? Ping @toji

[Mugen87]

I understand the motivation but continuously firing controllermove while in XR seems not right to me. As mentioned by @mrdoob in #20790 (comment), controllers in XR move all the time so it seems confusing to fire an event for something which is always happening.

I am trying to get XR controller input events, exactly like pointer and keyboard input.

I'm not sure this approach is right considering the difference in interaction between XR and classic mouse/keyboard and touch. To me, the concept of pointermove makes sense in this context but not in XR.

[arodic]

Realistically, don't vr controllers move all the time?

Yes @mrdoob, whenever they are being tracked. If turn off a controller or move it outside the tracking are it will stop emitting this event. This is similar behavior as pointer with stylus devices.

Without this event, XR controls would need to either:

1. Implement internal rAF(WebXR) to continuously read controller positions.
2. Rely on applications calling update() from within their own rAF.

If event mechanism is not an option, personally I lean towards former.

Closing due to no traction on this PR.

[mrdoob]

Now that I'm trying to get DAT.GUI working inside VR I can see why this is needed 😇

[mrdoob]

I think I'm going to rename it to move though.

[mrdoob]

Thanks!

[mrdoob]

Oh, hold on... Still not sure about this.
Don't we just need a single event being dispatched from controller?

[mrdoob]

@arodic does this work for you? a63ff05

[arodic]

I think it should work @mrdoob. Let me check...