yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> The problem is unshear also applies stretches, and it leads to very 
> complex corner cases (reason for the "cos" thing).
Whatever coordinates you use, spheres will transform to ellipsoids,
stretching the Aabb. As the transform stretches Aabb's too much (the
angle is too sharp), Aabb's sizes will approach the limit.

This is a theoretical question that you can't solve by saying "it
works"...

Send be the diff please (bzr diff), I will have a look ;-)

> Yes, I checked facets code too. No big deal. It is in fact simple to 
> strectch Aabbs.
Yes, but still won't let you do infinite shear. 
> > You have untransform sphere shape itself anyway, which gives ellipsoid,
> > in which case the Aabb is enlarged in the shear direction.

> Seriously, I see it running in front of me. You should trust me. :)
> The last problem is it is still not infinite shear, it stops at the same 
> point (ux/height=300% or so) than with unshear/size.
Depends what you run. The collider is quite complex thing and honestly I
don't even remember all codepaths the periodic code might take during
the sort; it took me weeks to get it right, so I am a little scared of
breaking it.

Come back later.

v