yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> No definitive answer on my side, but the "theoretical" answer could be : 
> simple shear can be seen as grains passing through a fixed rectangular 
> window (each time one grain goes out via the right hand side, it comes 
> back from the left).
> 
> If you see it that way, you easily realize that there is no physical 
> size restriction as in compression/tension/pure-shear.
OK, I see what you mean. You implemented what was already working,
sorry. (I was rotating the whole cell around about a week ago, with
matrix that was antisymmetric on non-diagonal.) You could've tried
before that it worked...

Whether you compute collisions on stretched or unstretched axes doesn't
make any difference to the algorithm. From my point of view, though, it
was cleaner to not stretch them, since you don't have to scale Aabb
backwards to compensate for that. (Cell::_size is computed once per
step, whereas you have to scale every single Aabb...)

v