I don't have any strong opinions regarding this (maybe I used to). I now think we should pick just one way to specify domains to avoid confusion.
-- Anders On Tue, Jun 03, 2014 at 12:58:51PM +0200, Martin Sandve Alnæs wrote: > No. The C++ Assembler does not use mesh.domains() but requires > the Form to have this data, and this is consistent with that behaviour. > > There's been a lot of discussions and strong opinions on > mesh.domains(), and unless we make a joint decision that > mesh.domains() should be used consistently in dolfin internals, > I'm simply ignoring the existence of mesh.domains() everywhere. > > Martin > > > On 3 June 2014 12:46, Joachim Berdal Haga <[email protected]> wrote: > > If the mesh contains subdomain markers (previously available via > mesh.domains().facet_data() or somesuch), is this automatically picked up > by ds(domain=mesh)? > > > > On 3 June 2014 11:41, Martin Sandve Alnæs <[email protected]> wrote: > > Attaching mesh functions to the form in legacy syntax looks like this: > > a = f*ds[facetfunction](1) > > or this: > > dss = ds[facetfunction] > b = f*dss(1) + g*dss(2) > > I would now recommend using the new syntax: > > a = f*ds(1, subdomain_data=facetfunction) > > dss = ds(subdomain_data=facetfunction) > b = f*dss(1) + g*dss(2) > > and passing the mesh as well is a good habit: > > dss = ds(domain=mesh, subdomain_data=facetfunction) > b = f*dss(1) + g*dss(2) > > because this will work even if f is an expression that doesn't contain > the mesh. > > Martin > > _______________________________________________ > fenics mailing list > [email protected] > http://fenicsproject.org/mailman/listinfo/fenics > > > > > > _______________________________________________ > fenics mailing list > [email protected] > http://fenicsproject.org/mailman/listinfo/fenics _______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
