Re: quaternions

[Pierre de Buyl]

Hi Felix,

On Wed, Dec 18, 2019 at 01:32:51PM +0100, Felix Höfling wrote:
> Very good idea. It is also pleasing to see that H5MD is alive - from daily
> experience it works great :-) And from recent discussions on a conference
> it seems that there is increased interest from more groups as well.

For my use cases, it "just works" and is a nice tool. I am glad that others use
it!

> Specific comments on the quaternions: The storage order (scalar part,
> vector part) follows also the notation I'm familiar with from the
> mathematical literature. And of course, the big packages leave no other
> choice.

Indeed. I should have checked the common usage before writing my code :-)

> Where should it go? I suppose it would be listed as a property in the
> particles/ tree, next to "position"?

The plan is to have it as "an element in the particles group", so that it would
be next to position.

> My own HALMD software is close to support quaternions as well. Here a
> question came up, which we may address now: What is the relation of the
> quaternion to the axis vector for axisymmetric particles (e.g.,
> ellipsoids)? A natural name for both properties would be "orientation" -
> so how should we distinguish them in H5MD? Of course, we could call the
> quaternion representation "orientation" and the axis representation "axis"
> or "direction" or so. Any thoughts on this?

No thought yet. It is true that for axisymmetric bodies, quaternions are not
necessary. For quaternions, I see no issue with the name "quaternion", it is
unambiguous and actually used internally (even though "q" or "quat" would be the
variable name, the data is named quaternion in the code and documentation such
as "the quaternion" as one would say "the position").

What you refer to as axis is the unit vector pointing along the "long" axis of
the ellipsoid, right? If so, I can think of "axis", "u", "unit_vector", or even
"orientation" actually. All current names in the particles group (even the new
quaternion) are unambiguous. Apart from "id", we even have full english names
for all elements :-)

Do you agree to keep "quaternion" for the quaternion? If so, we are left with a
name to find for the axis.

Regards,

Pierre