ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Wed, Aug 31, 2005 at 06:22:15PM +0200, Garth N. Wells wrote:
> 
> Quoting Anders Logg <logg@xxxxxxxxx>:
> 
> > Of course, FFC is still very experimental but I don't think it would
> > be a good idea to base the FFC algebra on Numeric.array.
> > 
> > 1. I think the best solution to Garth's problem would be to let the
> > constructor of an Index accept one optional extra argument to allow
> > specification of ranges. Then one can do stuff like
> > 
> >     i = Index(0, d)
> >     j = Index(0, d)
> >     a = u[i].dx(i) * v[j].dx(j)
> > 
> > This would force corresponding Indices to match and I think it would
> > allow Garth to do what he likes. Correct me if I'm wrong.
> 
> This would be good.

Fine. I'll get to it as soon as I can.

/Anders