[feat] allow GLBufferAttribute as an attribute type

[mikkom]

This is the same change as #295, just retargeted against dev branch.

Why

The current typings don't allow setting GLBufferAttribute as an attribute in BufferGeometry. This functionality has been in Three.js for quite some time (originally introduced in mrdoob/three.js#13196 back in 2020).

Three.js master continues to support this feature in BufferGeometry (see e.g. https://github.com/mrdoob/three.js/blob/master/src/core/BufferGeometry.js#L371), documents the attribute type in https://threejs.org/docs/#api/en/core/GLBufferAttribute and has a usage example in https://github.com/mrdoob/three.js/blob/master/examples/webgl_buffergeometry_glbufferattribute.html#L115-L117.

What

This change adds GLBufferAttribute as a valid type for setAttribute and getAttribute methods in BufferGeometry. The type for GLBufferAttribute had already been defined and exported but it was not used anywhere else in the typings.

Note that the change can break consumers of the typings as GLBufferAttribute does not implement the same interface as BufferAttribute. The way to fix these issues is to use the boolean flags (isBufferAttribute, isInterleavedBufferAttribute, or isGLBufferAttribute) to discriminate between the different options (or alternatively by resorting to type assertions). Despite the potential breakage, I believe these typings are more correct.

Checklist

• Checked the target branch (current goes master, next goes dev)
• Added myself to contributors table
• Ready to be merged

[0b5vr]

oh, thanks for retargeting the target branch

[0b5vr]

if it's undefined, we don't have to explicitly say this I think.

[0b5vr]

ah, I just saw the message you left in the previous PR. #295 (comment)

If it's the solution I feel like we have to set isXXX to every Three.js class, and I don't think it's the right way.
(e.g. saying readonly isMesh: undefined in Matrix4 because someone might want to use that)
I believe people solve this kind of problems in other ways in the TypeScript realm.

[mikkom]

Sorry about the delayed response. I think you make a fair point there. It could indeed get unwieldy quite fast if we would need to cover all the fields in similar fashion. So I've now removed the explicit undefined field definitions and switched to using instanceof in the pointercontrols test.

[0b5vr]

Now it looks good, Thanks!