dolfin team mailing list archive

[Marie Rognes <meg@xxxxxxxxx>]

Peter Brune wrote:
> Ah, let me explain further: The reason I have the two vector spaces -- 
> affine and iso, is so that I can reverse the affine transformation 
> before applying the isoparametric transformation.  Today I found a 
> small bug in my isoparametric transformer, and now I'm sure stokes 
> works; the pretty example showing the differences between refined 
> affine and unrefined isoparametric is coming along!  We want to get 
> rid of having to transform back to the reference using the affine 
> coefficient space before this is ready to be used.  I have been 
> looking through UFL and FFC for a good way to drop in a new jacobian.  
> Passing the new jacobian to FFC is really easy using metadata... using 
> it is harder.
>
> So, anyone else have any tips on that?  Here's the MAJOR issue I can 
> see cropping up -- I can pepper the form itself with the new Jacobian 
> in order to transform function (derivatives) affinely, but other 
> transformations (co/contravariant Piola anyone?) will not work like this.

If you can do the stiffness matrix, you should also be able to do 
co-/contravariant Piolas?

--
Marie