yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> I was about to send questions about this a few days ago.
> Do you mean one-side multiplication returns the same result as two-side 
> for normalized quaternions? Did you check that?
Eh, no. I think in mathematic formal notation, you ahve to multiply from
both sides, but the c++ interface is such that you multiply only from
the left (always). 

Here
http://en.wikipedia.org/wiki/Quaternions_and_spatial_rotation#Describing_rotations_with_quaternions they have the vector actually represented as quaternion with the real part zero, then multiplying from both sides... To me it is not very clear, though, but I know that it works fine in Yade (Dem3DofGeom code uses Quaternions quite a lot).

> Can we consider quaternions are always normalized?
Yes, in practice. To avoid numeric drift, they are normalized once each
step in NewtonIntegrator (leapFrogSphericalRotate), so we are always
sufficiently close to normal quaternions (which do pure rotation,
without distoring the space).

v.