ufl team mailing list archive

[Mehdi Nikbakht <m.nikbakht@xxxxxxxxxx>]

On Wed, 2010-04-21 at 16:18 +0200, Kristian Oelgaard wrote:
> 
> On 21 April 2010 12:09, Mehdi Nikbakht <m.nikbakht@xxxxxxxxxx> wrote:
> > Hello,
> >
> > I am just started examining ufl to expand integrands. To perform the
> > expansion, I have defined a transformation derived form
> > ReusedTransformer. I succeeded expanding ufl integrands for scaler
> > functions, but I encountered a problem while trying to work with vector
> > functions.
> >
> > Assume that I have a linear form defined as,
> >
> > element = VectorElement("Lagrange", triangle, 1)
> >
> > u = TrialFunction(element)
> > v = TestFunction(element)
> >
> > f = Coefficient(element)
> > g = Coefficient(element)
> >
> > L = inner(f + g, v)*dx
> >
> > which will be interpersonal as,
> > sum_{i_0} ((v_0)[i_0]) * ((w_0 + w_1)[i_0])
> >
> > I want to use ufl to expand linear form to,
> >
> > L = inner(f, v)*dx + inner(g, v)*dx
> > which should be interpreted as,
> >
> > sum_{i_0} ((v_0)[i_0]) * ((w_0)[i_0]) + sum_{i_1} ((v_0)[i_1]) *
> > (( w_1)[i_1])
> >
> > Now I am getting
> > sum_{i_0} ((v_0)[i_0]) * ((w_0)[i_0]) + sum_{i_0} ((v_0)[i_0]) *
> > (( w_1)[i_0])
> >
> > which fails by an error message as,
> >
> > FAILURE in reuse_if_possible:
> > type(o) = <class 'ufl.indexsum.IndexSum'>
> > operands =
> >
> > sum_{i_0} ((v_0)[i_0]) * ((w_0)[i_0])  + sum_{i_0} ((v_0)[i_0]) *
> > ((w_1)[i_0])
> >
> >
> > It seems to perform this extension, I need to update indices defined for
> > summation. My question is how I can update an index like i_0 to i_1?
> 
> I don't see why you need one of the indices to be i_1, the above code seems to make just as much sense.
> I think there is something else going wrong in your transformer class.

If we compile,
 L = inner(f, v)*dx + inner(g, v)*dx

the ufl integrand is,

sum_{i_0} ((v_0)[i_0]) * ((w_0)[i_0]) + sum_{i_1} ((v_0)[i_1]) *
 (( w_1)[i_1])

although having the same indices for summation is valid for these two
terms in integrand in index notation, but it seems that it is not case
for ufl.

Mehdi
> 
> Kristian
> 
> > I hope that my question was clear enough.
> > Mehdi
> >
> > _______________________________________________
> > Mailing list: https://launchpad.net/~ufl
> > Post to     : ufl@xxxxxxxxxxxxxxxxxxx
> > Unsubscribe : https://launchpad.net/~ufl
> > More help   : https://help.launchpad.net/ListHelp
> >
>