yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> > But I get in trouble there is no easy way to separate scaling and
> > (rotation+shear), right? Setting 1s on the diagonal is clearly nonsense,
> > although that is what we (I) do now.
> >
> >   
> I wish I could answer, but I have no clear and general solution. For 
> instance, I browsed OGL doc yesterday and I still don't know if scale 
> and shear can be done in only one operation using F and glMultMatrixd. 
> Actually, I don't even understand how the current 4x4 matrix can shear 
> the box correctly... Do you know a good web page about this? 
> http://www.songho.ca/opengl/gl_transform.html#matrix doesn't work.
> 
> For display, at least, one solution would be to draw the 12 edges one by 
> one (origin->Hsize[0],origin->Hsize[1],etc.). (note : there is commented 
> code in periController to recompute Hsize with refSize and F.)
> 
> Perhaps list the precise problems here, so we can think about them. For 
> the most general answer on scaling/rotation split, I would think in 
> terms of R.U polar 
> factorisation(http://en.wikipedia.org/wiki/Finite_strain_theory), but 
> I'm not sure we really need that.
I think I am on the right track now with this. Not as difficult as I
tought.