On Thu, Dec 07, 2006 at 01:18:49AM +0100, angelo simone wrote:
Here's the bundle. dolfin-convert is now in pure python (i.e. there are
no calls to grep) and there are no temporary files. Is it ok now?
The script has been tested with the attached mesh files:
gmsh.msh (2d)
old-xml.xml.gz (2d)
tetgen.mesh.gz (3d)
Thanks, but I don't really like the write_output stuff. I have added
your changes and modified mesh2xml so it first scans the file for the
cell type and then does the conversion. Could you supply a patch that
does the same thing for gmsh2xml and xml_old2xml?
I have tried the output of the first two conversions with a Dolfin
example and they worked out fine. The tetgen mesh, which comes from the
example.poly file distributed with tetgen 1.4.1, has not been tested
against Dolfin. The output does look ok though.
Thanks. I have added the test meshes and your updated conversion
script, but I don't like the write_output stuff
As far as I can see from the tetgen web page
(http://tetgen.berlios.de/), tetgen cannot generate 2d meshes. Is there
a reason why in the script there is a check on num_dims == 2?
I haven't tested, but presumably the conversion works for meshes
generated by Triangle. I think Triangle and TetGen share the same format.
I will soon send a bundle to update section 8.3 of the
dolfin-user-manual. A bundle on the procedure to create a bundle will be
ready soon as well.
Very nice.
/Anders
angelo
Anders Logg wrote:
You could try generating some with tetgen.
When you submit the patch, please also include some (very small) test
meshes that you have run the code on. We can add these for future
reference.
Thanks
/Anders
On Tue, Dec 05, 2006 at 01:56:00AM +0100, angelo simone wrote:
Exactly. Anyway, I got rid of the ugly hack... ;-)
Before sending a bundle, I would like to test the new version on 2d
and 3d Medit meshes, which I do not have. Does anyone have them?
Thanks!
angelo
Anders Logg wrote:
I think because the new mesh format requires celltype in the header of
mesh, so the mesh needs to be scanned before writing to find out the
celltype.
Instead of doing this, the patch first generates an illegal xml file
missing the celltype attribute and then scans the generated file to
grep for the cell type and then fixes the header...
It works but it's not a nice solution. Angelo, please submit an
updated patch that first scans the file for the cell type (preferably
not using grep).
/Anders
PS: Yes, I know you looked at xml_old2xml() that I wrote and I do use
grep, but it's still an ugly hack. :-)
On Mon, Dec 04, 2006 at 11:06:21AM +0100, Garth N. Wells wrote:
Why are temporary files being created? Isn't updating to the new
mesh xml format just a case of changing the output formatting?
Garth
angelo simone wrote:
hi,
I have fixed dolfin-convert (patch attached). Now the output is in
the current format for both Gmsh and Medit. I have tested
dolfin-convert with triangular Gmsh meshes but is should work fine
with all the remaining combinations as well.
Regarding the addition to the user manual, maybe we should just add
that on top of the steps described in the Gmsh mesh generation
tutorial
(http://www.geuz.org/gmsh/doc/gui_tutorial/gui_tutorial.pdf) one
needs to create physical surfaces/volumes as well to generate a
dolfin-friendly mesh -- yesterday I have used Gmsh for the first
time... that's why I did that mistake. Anyway, dolfin-convert
complains if the mesh is not in the correct format.
angelo
