ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Nov 06, 2008 at 04:24:06PM +0000, Garth N. Wells wrote:
> 
> 
> Anders Logg wrote:
> > On Thu, Nov 06, 2008 at 11:21:02AM +0100, Kristian Oelgaard wrote:
> >> Quoting Anders Logg <logg@xxxxxxxxx>:
> >>
> >>> On Wed, Nov 05, 2008 at 03:17:13PM +0000, Garth N. Wells wrote:
> >>>> In the FFC-produced wrapper code for DOLFIN, could will still generate a 
> >>>> Form constructor which takes the coefficient functions (or a vector of 
> >>>> coefficient functions) as an argument? This is needed when one works 
> >>>> with a pointer to a form. Doing
> >>>>
> >>>>    Form* a;
> >>>>    a = new WeightedPoissonBilinearForm(V, V);
> >>>>
> >>>> it's not possible to attach coefficient functions because a Form doesn't 
> >>>> know about them, only WeightedPoissonBilinearForm does.
> >>>>
> >>>> Garth
> >>> Is this something that happens in user code (so it needs to look
> >>> nice)?
> >>>
> >>> Otherwise, just do
> >>>
> >>>   Form* a;
> >>>   static_cast<WeightedPoissonBilinearForm*>(a)->f = f;
> >> In this case, how and when do you set the FunctionSpace of the form?
> >> I don't see a problem in having the old constructor of the type
> >>
> >> PoissonLinearForm(dolfin::FunctionSpace& V0, dolfin::Function& v0)
> >>
> >> existing alongside the new approach such that the user can avoid doing too much
> >> complicated stuff.
> >>
> >> Kristian
> > 
> > I agree. I'll add it back again.
> >
> 
> Could you take a look at the wrapper code for Functional too while 
> you're at it? Presently, two identical constructors are being produced 
> by FFC.
> 
> Garth

ok.

-- 
Anders

Attachment: signature.asc
Description: Digital signature