ffc team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

Martin Sandve Alnæs wrote:
> 2007/10/12, Anders Logg <logg@xxxxxxxxx>:
> > Martin Sandve Alnæs wrote:
> > > 2007/10/12, Martin Sandve Alnæs <martinal@xxxxxxxxx>:
> > > > 1) What is wrong with this form?
> > You take a dot product of a scalar with a vector (or a vector with a
> > matrix in the vector case).
>
> While I agree of course that the scalar case is wrong, I'd expect dot
> to handle vector times matrix, the convection term is usually written
> "u dot grad u".

Yes, we need to fix the dot operator.

In the meantime, do

  u[j]*D(u[i], j)

or

  w[j]*D(u[i], j)

/Anders

> > > > 2) If it's a user mistake and not an ffc bug, ffc should raise an
> > > > exception and not return None
> > Yes. Looking at operators.py, the "else" clause seems to be missing for
> > the dot function.
> >
> > /Anders
> >
> >
> > > > from ffc import *
> > > >
> > > > fe = FiniteElement("Lagrange", "triangle", 1)
> > > This should be:
> > > fe = VectorElement("Lagrange", "triangle", 1)
> > > but it doesn't change the None returned from dot.
> > >
> > > > v  = TestFunction(fe)
> > > > u  = TrialFunction(fe)
> > > > w  = Function(fe)
> > > >
> > > > Dw = grad(w)
> > > > Du = grad(u)
> > > >
> > > > uDw = dot(u, Dw)
> > > >
> > > > print u
> > > > print Dw
> > > > print uDw # None!
> > > Output:
> > >
> > > vi-1
> > > [[w0_a9(dXa11/dx0) | ((d/dXa11)va9[0]), w0_a12(dXa14/dx1) |
> > > ((d/dXa14)va12[0])], [w0_a15(dXa17/dx0) | ((d/dXa17)va15[1]),
> > > w0_a18(dXa20/dx1) | ((d/dXa20)va18[1])]]
> > > None
> > >
> > >
> > > > wDu = dot(w, Du)
> > > >
> > > > a = dot(v, uDw+wDu) * dx
> > > >
> > > >
> > > >
> > > > --
> > > > Martin