ufl team mailing list archive

[Kent-Andre Mardal <kent-and@xxxxxxxxx>]

man, 14.01.2008 kl. 23.37 +0100, skrev Martin Sandve Alnæs:
> 2008/1/14, Anders Logg <logg@xxxxxxxxx>:
> > On Mon, Jan 14, 2008 at 09:31:03PM +0100, Martin Sandve Alnæs wrote:
> > > I suggest adding dyadic notation to UFL as well, something like:
> > >
> > > ei(i) -> unit vector in direction i
> > > eij(i,j) -> unit 2nd order tensor in direction i,j
> > >
> > > i.e. eij(i,j) == outer(ei(i), ej(j))
> > >
> > > # defining a vector by adding components:
> > > uu = u*ei(0) + v*ei(1) + w*ei(2)
> > >
> > > # defining a diagonal matrix:
> > > sigma_0 = T * (a*eii(0,0) + b*eii(1,1) + c*eii(2,2))
> > >
> > > # defining a direction vector from an angle:
> > > d = cos(theta)*ei(0) + sin(theta)*ei(1)
> >
> > Sounds good.
> >
> > How does one write the identity matrix?
> >
> > --
> > Anders
> 
> Either one of these should work:
>   dij = dot(e(i), e(j)) # Kronecker delta
>   I = outer(ei(i), ei(i)) # sum of outer products of ei(i) with
> itself, ie sum of eij(i,i) for all i
> 
> But we can still have "I = Id()" or similar for that one.
> 
> (Note that I don't see all the details of the index notation
> implementation clearly yet).
> 

Great, we also need this (or something similar) if we implement the
exterior calculus with UFL. 

Kent