![]() Mesh.save subdomains and boundaries #261 I plan to deprecate our in-house JSON format and try to use meshio instead.But then the problem might be that either meshio won't reread those files as having subdomains or boundaries or won't put them into the (unfortunately named) 'gmsh:physical' entries which will mean that they aren't recognized by _meshio. So I'm wondering whether it might be better to avoid that conversion and see whether (already or in the future) meshio can hold and write something closer to. But when Gmsh suddenly started generating MSH 4.1 by default which nothing but itself could read, I hastily added a facility to meshio to read it, but to more or less coerce it into the same data-structure as previously used for MSH 2, This involves conversion from the second representation to the first, lossily, which leads to problems like nschloe/meshio#524 when subdomains or boundaries aren't mutually exclusive (something which makes sense in 'multiphysics' problems). The current behaviour of meshio.read is mostly the way it is because at the time it was developed, the only format in which meshes were generated with subdomains and boundaries was Gmsh's MSH 2 and that used the, although I haven't fully understood its concept of 'entity'. boundaries attributes of a skfem.Mesh are different they're Dict, with the terms of the ndarray values being indices of meshio_type or bnd_type belonging to the subdomain or boundary.Īre discussed in nschloe/meshio#175, nschloe/meshio#290, nschloe/meshio#347, … These are interpreted as one-dimensional integer arrays of length equal to the number of meshio_type and bnd_type cells specified in the. The current looks for a key 'gmsh:physical' in the meshio_type and bnd_type entries of the. Not a huge task to implement, no, but maybe there is a preliminary question to think about: ‘tagged subsets of elements’ nschloe/meshio#290.
0 Comments
Leave a Reply. |