David, the code snippet you posted is not going to work. You can't use a FEValues without initializing it with a cell_iterator. If all the information you need for your postprocessing is contained in a data vector based on a single DoFHandler, then DataPostprocessor is what you want to use. If you take a look at the interface at the signature of DataPostprocessor<dim>::compute_derived_quantities_scalar[1], you see that this provides the access you want. Namely, you access at the vertices of the patches used for output.
Best, Daniel [1] https://www.dealii.org/8.4.0/doxygen/deal.II/classDataPostprocessor.html#af8f60ada7b7ae5417923613d87e16e0d Am Sonntag, 22. Mai 2016 14:45:55 UTC+2 schrieb David: > > I decided to use DoFTool::map_dofs_to_support_points(), so my code above > will be like this: > > DoFTools::map_dofs_to_support_points(MappingQ1<dim>, dof_handler, > std::vector<Point<dim>> all_nodes); > std::vector<double> principle_stress; > principle_stress.resize(dof_handler.n_dofs()); //suppose dofs = > no_of_nodes here. > for (unsigned int i = 0; i != dof_handler.n_dofs(); ++i) > { > Point<dim> p = all_nodes[i]; > Quadrature<dim> quad(p); > FEValues<dim> fe_values (fe, quad, |update_gradient); > > ... ... //compute stress with fe_values.shape_grad(i, q) > principle_stress[i] = ... ... > } > > //output stress results > DataOut<dim> data_out; > data_out.attach_dof_handler(dof_handler); > data_out.add_data_vector(principle_stress,"principle stress") > > But here I'm not sure whether the ordering of points in the vector > "all_nodes" getting from using DoFTool::map_dofs_to_support_points() is the > same with dof_indices of the dof_handler. If so, there is no problem with > the above code. While if not so, the ordering of nodal stress solution > would also be different from dof_handler, as a result, some stress is > attached to wrong dofs by DataOut. > > So which case is correct, could anyone give some explanations? > > Great thanks! > > Best, > David. > > On Sunday, May 22, 2016 at 7:43:38 PM UTC+8, David wrote: >> >> Hi, Daniel: >> >> Thank you for your help. But this seems not to be convenient. I am >> thinking to run my code like the following (psuedocode) >> >> for (i=0; i != number_of_nodes; ++i){ >> Point<dim> pi = get_point_from_global_index(i); >> Quadrature quad(pi); >> FEValues<dim> fe_values (fe, quad, |update_gradients); >> >> ... ...//do some computations with fe_values.shape_grad(i, q) >> } >> >> In this way, I can easily run over all the triangulation nodes without >> worrying about worrying about recomputation at some nodes if I instead run >> over each cell. >> But the difficullty here that I can't find a function which can do the >> same job as "get_point_from_global_index(i)", namely give me the point with >> global node index i of the triangulation. >> >> Is there any function in dealii with this functionality? >> >> Great thanks! >> >> Best, >> David >> > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
