yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

> I was actually thinking about implementing contact geometry classes from
> scratch, both incremental and non-incremental versions, in such way that
> deformations (normal, shear, twist, bending) would be precomputed in the
> Ig2_* functor, and then just accessed directly from the constitutive law.
>   
It would make sense. I don't remember why the updateShearForce code was 
moved to interaction geometry. Rotating the force in geometry is ok, but 
computing the shear force increment should be done in ElasticContactLaw.
There has been a bit of over-splitting actually, since the incremental 
ElasticContactLaw can fit in almost 15 lines of code in total 
(geometry+physics+constitutiveLaw). I don't think it will change anytime 
soon though.

> The InteractionGeometry classes themselves would be the same, only the
> functor would make the distinction of absolute/incremental. There already
> was 6dof contect implemented (Dem6Dof_Sphere_Sphere), but it is commented
> out currently; and guess what, it is much cleaner code than the horrible
> mess (sorry) in CohesiveFrictionalContactLaw.
>
>   
I though I would start from this Dem6Dof next time I have to implement a 
new law. But you know, it is always the same problem : I know the 
CohesiveFrictional code and I tested it. Learning new notations will 
take time,  while it needs only few minutes to do what I want with the 
old code. At least, I'll recommend new users to start with Dem6Dof if 
they want to derive a new constitutive law (which is not what they seem 
to do unfortunately).

"Mess" is a relative concept : it took me approximately the same time to 
understand Galizzi's SortCollider and Smilauer's one. I'm not sure which 
one is better. Well, I tend to think yours is, but the biggest part is 
different coding styles from different developpers.
I'm not trying to say CohesiveFrictionalContactLaw is well written 
though. It was not very well structured from the very beginnning, and 
Janek added tests for snow creep, which doesn't help. I'll clean that 
eventually, but the code can't be simpler than the algorithm.

Bruno