pex-gl
Confusion between transform.matrix and transform.position/rotation/scale
Currently a transform component is the combination of position/rotation/scale multiplied by matrix if it exists.
Any reason to store this extra matrix component (or vice-versa position/rotation/scale)? The orbiter doesn't account for it. As a result, it is not possible to have only transform.matrix to define an object transformation (eg. a camera) and update it with an orbiter.
Nowhere in pex-renderer codebase the transform.matrix is used if I am correct.
Its required for gltf compatibility. We might try matrix decomposition but i had some problems with that in the past.
Are gltf cameras transforms matrices?
glTF cameras are nodes like any other, so they can have either translation/rotation/scale or matrix.
I'd be more inclined to have only one set of local transformation information per transform but you're saying the following can cause issues?
const transform = node.matrix
? {
position: [
node.matrix[12],
node.matrix[13],
node.matrix[14]
],
rotation: quat.fromMat4(quat.create(), node.matrix),
scale: [
Math.hypot(node.matrix[0], node.matrix[1], node.matrix[2]),
Math.hypot(node.matrix[4], node.matrix[5], node.matrix[6]),
Math.hypot(node.matrix[8], node.matrix[9], node.matrix[10])
]
}
: {
position: node.translation || [0, 0, 0],
rotation: node.rotation || [0, 0, 0, 1],
scale: node.scale || [1, 1, 1]
}
Seems like the glTF2.0 Spec mandates that decomposition is always possible.
Matrices must be decomposable to TRS. This implies that transformation matrices cannot skew or shear.
@damien For the record: where is your code from?
Todo:
- [ ] implement above change and remove node.matrix
- [x] check glTF examples
Our source: https://github.com/pex-gl/pex-renderer/blob/master/loaders/glTF.js#L529 Reference from: https://github.com/KhronosGroup/glTF-Sample-Viewer/blob/master/src/node.js#L58
Important fix they did: https://github.com/KhronosGroup/glTF-Sample-Viewer/commit/89045ce125d989f55c21a69a06e987730745b3fa