dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Apr 05, 2007 at 02:07:27PM +0200, Garth N. Wells wrote:
> The code in ConstantFunction::interpolate does look right to me. 
> Shouldn't it just return "value"?
> 
> Garth

Yes, in simple cases where dofs are just point evaluation, but in
general the values of the dofs may differ from the point values.

For example, a BDM1 element has dofs given by point evaluation (at
certain points) of the normal components and the normal component of a
vector-valued function at a point on the edge of a triangle may be
different from its point value at that point.

Currently, FFC is only able to generate evaluate_dof() for Lagrange
and vector-valued Lagrange, so right now, the extra work of going
through evaluate_dof does not buy us anything more than just putting
"value" into "coefficients". (But this will change when evaluate_dof()
gets implemented for other elements in FFC.)

/Anders