Fix 2D position computation

[eve-le-guillou]

Hi all,

This PR fixes an incoherence in triangleToPosition2D.
Originally, it would output the position in terms of triangles (and not VTK squares).
The problem is that its 3D counterpart tetrahedronToPosition, outputs the number of VTK cubes. This creates an incoherence when both methods are used within the same code.

Dividing p[0] in triangleToPosition2D makes the method output squares. However, this means changing as well getCellVertex, so that it does not divide by two for the 2D case.

This was detected due to the use of both functions in MPI features.

To test that these changes don't impact more of the code, I computed the DiscreteGradient on the cells.vti dataset. No differences have been detected when comparing the results with and without the changes.

Thanks for any feedback,

[julien-tierny]

thanks a lot @eve-le-guillou !
it looks all good to me.
@pierre-guillou are you OK with this?

[pierre-guillou]

You might also need to modify the ImplicitTriangulationCRTP::getTriangleEdgeInternal method, playing with the TriangulationRequest filter with Simplex "Triangle" and Request Type "Facet" outputs wrong edges.

The ImplicitTriangulationCRTP::getTriangleNeighborNumber and ImplicitTriangulationCRTP::getTriangleNeighbor methods have similar p[0] / 2 expressions, they might need additional correction too.

[eve-le-guillou]

Hi,

I added the recommended changes and also changed the corresponding methods in the periodic triangulation (that also showed some /2 on the 0 coordinate in 2D).

It should be good now.