From: Martin <luthi@gi...>  20050413 23:05:39

Roy Thanks for your answers. Roy Stogner writes: > On Tue, 12 Apr 2005, Martin L=FCthi wrote: >=20 > > Maybe I am doing something wrong, any help appreciated. >=20 > Not wrong so much as incompatible. The write_nodal_data function > expects a vector whose indices correspond to node number. The > project_vector function handles vectors whose indices correspond to > global degree of freedom numbers. The latter is more useful  > generally you want meshdependent data to be represented by a finite= > element space, which may not be piecewise linear. Yup, I see.=20 > The problem is that we don't have a MeshOutput method for plotting > such a vector. When I've needed one output it was in a singlesyste= m > problem and so I could work around that deficiency: >=20 > NumericVector<Number> &old_solution =3D system.get_vector("old_so= lution"); > //... > (*system.solution) =3D old_solution; > system.update(); > GMVIO(es.get_mesh()).write_equation_systems > (output_file.str(), es); Following prior advice I set up a dummy equation system and had the vector values stored as the solution there. Also quite a bit of overhead. > That code kind of sucks. Any idea what the new method ought to be > named? write_global_degree_of_freedom_data() doesn't have a nice ri= ng write_vector(); ? Best, Martin 