@gertjanvanzwieten Could you please allow me to push to your branch. That would make things easier. Thanks!

@uekerman try now? I think you should have full rights.

@gertjanvanzwieten Works. Thanks!
(When opening a PR you can tick "allow push access by maintainers" or similar. Then, you don't have to give access to your complete fork, but only the branch. But I don't complain 😊 )

Ready for review.

• @MakisH Could you please check for tutorials style
• @IshaanDesai Could you please check the Nutils codes and run them

Motivation for this case is to showcase the direct mesh access feature, developed by @davidscn. We will add more participants here in the future (potentially a G+Smo one).

Could not be merged with the existing partitioned-heat-conduction tutorial as we need a different preCICE config.

I also have a higher-order solver using the direct-mesh access with this tutorial. I will also check for the convergence behavior when coupling with the Nutils code here.

Tested it and it works like charm, even the high-order properties are preserved (as expected).

Tested it and it works like charm, even the high-order properties are preserved (as expected).

What kind of tests did you do? Also different meshes left and right?

What kind of tests did you do? Also different meshes left and right?

deal.II uses hexahedral meshes and Nutils tet meshes, so I'm not sure what exactly you want to have more different. Maybe I should also add that I didn't do a full convergence study here, but one can clearly see the error drop when going from first order polynomials to seconds order polynomials. I can test things more thoroughly if there is any need. Right now, it was more or less meant as a consistency check.

Edit: Only thing which looks a bit confusing is the reported number of DoFs by Nutils vs the vtk output. Are higher degrees sub-sampled in the vtk output?

Only thing which looks a bit confusing is the reported number of DoFs by Nutils vs the vtk output. Are higher degrees sub-sampled in the vtk output?

Exactly, this happens here:

bezier = domain.sample('bezier', degree * 2)
...
x, u, uexact = bezier.eval(['x_i', 'u', 'uexact'] @ ns, lhs=lhs, t=t)

@MakisH Ready to merge from my side