Conversion between euler and quats doesn't seem stable

[rvanlaar]

While creating a pygfx Controller I noticed the following:

Moving from euler to quats and back is not stable.
Here's an example:

In [11]: la.quat_to_euler(la.quat_from_euler(np.array([math.pi/2, math.pi/180,0])))
Out[11]: array([1.57079633e+00, 1.73472348e-18, 1.74532925e-02])

Almost as if the last two items of the array are reversed.

In code:

import math
import numpy as np
import pylinalg as la

la.quat_to_euler(la.quat_from_euler(np.array([math.pi/2,math.pi/180,0])))

[rvanlaar]

Looked at scipy, and that one works:

import numpy as np
import pylinalg as la

from scipy.spatial.transform import Rotation


quat = np.array([0.70707986, 0.00617059, 0.00617059, 0.70707986])

la.quat_to_euler(quat)
Rotation.from_quat(quat).as_euler("XYZ")

[Korijn]

Would you mind opening a pull request with a unit test that demonstrates the failure?

[rvanlaar]

Done: #87

It seems the issue has to do with the order. The default uppercase "XYZ" has a different result than lowercase "xyz".

[Korijn]

pylinalg/pylinalg/quaternion.py

Line 296 in 92a83a2

is_extrinsic = [x.isupper() for x in order]

Using uppercase letters apparently makes that axis extrensic. I don't really understand what that means.

[rvanlaar]

The three rotations can either be in a global frame of reference (extrinsic) or in a 
body centred frame of reference (intrinsic), which is attached to, and moves with, the object under rotation 

https://docs.scipy.org/doc/scipy/reference/generated/scipy.spatial.transform.Rotation.from_euler.html

My 2cts:

• Document the behavior
• give quat_to_euler also an order
• have quat_to_euler and quat_from_euler use the same default order.

[Korijn]

If you want to take a shot at it, you're more than welcome. We're shorthanded at the moment.

[rvanlaar]

Thanks for the offer. Am very busy, but who knows.