On 04/27/10 02:52, Andre Smit wrote:

http://code.google.com/p/sfepy/wiki/Primer

Rough draft - please edit as you see fit. Comments and criticism welcome :)

Great job!

I have just a few comments:

• to prevent the wiki engine to think that SomeWord is a link, it is possible

to write it as !SomeWord. So maybe you want change SfePy -> !SfePy, NumPy -> !Numpy etc.

• one could alternatively use problem.solve() in place of eval_term_op() in

stress_strain().

• it is advisable to weight the contributions of each node in nodal_stress() by

the areas (volumes in 3D) of the elements it is in. In fact, Variable.data_from_qp() would do that for you! Yes, another undocumented function :].

• the zeros in iconn correspond to element groups - an element is referred to

by (ig, iel) - the group number and its index within a group. You get different groups either when having both triangles and rectangles in one mesh, and/or by using several element groups ids in the mesh file - for medit meshes it is the last number in each row of the element connectivity. This number corresponds to the one in 'elements of group <integer>' region selector.

I have also added a link to the sfepy.mechanics.units module and removed a duplicate word.

Thanks for writing the primer! You discovered a bunch of bugs in the process, so it lead to improved code too.

cheers, r.