Adapted most GridapEmbeddedTests and examples to Gridap v0.15

[ericneiva]

Hi @fverdugo,

I have adapted most GridapEmbedded tests to Gridap v0.15:

There is a couple of issues, before proceeding with the PR, though:

1. StokesCutFEMTests.jl: I cannot compute this integral (I get a "This function is not yet implemented").

  ...
  Γi = SkeletonTriangulation(cutgeo_facets)
  n_Γi = get_normal_vector(Γi)
  ...
  dΓi = LebesgueMeasure(Γi,degree)
  ...
  c_Γi(p,q) = (β0*h)*jump(p)*jump(q)
  a((u,p),(v,q)) = ∫( - c_Γi(p,q) ) * dΓi

See also this line in StokesCutFEMTests.jl. What could be going wrong? Am I missing something?

2. BimaterialPoissonCutFEMTests.jl: I had to change how to compute the cell-dependent weights and the stabilisation param to make things work, otherwise I get the array dimension mismatch assertion:

You are trying to build a CellField from an array of length 142
on a Triangulation with 1800 cells. The length of the given array
and the number of cells should match.

when attempting to integrate the weak form:

jump_u(u1,u2) = u1 - u2
mean_q(u1,u2) = κ1*α1*∇(u1) + κ2*α2*∇(u2)
mean_u(u1,u2) = κ2*u1 + κ1*u2
a((u1,u2),(v1,v2)) =
  ∫( α1*∇(v1)⋅∇(u1) ) * dΩ1 + ∫( α2*∇(v2)⋅∇(u2) ) * dΩ2 +
  ∫( β*jump_u(v1,v2)*jump_u(u1,u2)
     - n_Γ⋅mean_q(u1,u2)*jump_u(v1,v2)
     - n_Γ⋅mean_q(v1,v2)*jump_u(u1,u2) ) * dΓ

Now, I compute them at all cells to match the size.

  const κ1 = (α2*meas_K1) ./ (α2*meas_K1 .+ α1*meas_K2)
  const κ2 = (α1*meas_K2) ./ (α2*meas_K1 .+ α1*meas_K2)
  const β = (γ_hat*meas_KΓ) ./ ( meas_K1/α1 .+ meas_K2/α2 )

Before, we only needed the ones at the cut (142).

  const κ1 = (α2*meas_K1_Γ) ./ (α2*meas_K1_Γ .+ α1*meas_K2_Γ)
  const κ2 = (α1*meas_K2_Γ) ./ (α2*meas_K1_Γ .+ α1*meas_K2_Γ)
  const β = (γ_hat*meas_KΓ_Γ) ./ ( meas_K1_Γ/α1 .+ meas_K2_Γ/α2 )

Is that alright? Or, again. am I missing something?

[fverdugo]

1. It looks like a bug. I'll take care.

2. The error you get about the number of cells mismatch is the expected behavior. Your solution looks OK.

[ericneiva]

1. It looks like a bug. I'll take care.

Okay, let me now when it's solved. Thanks.

2. The error you get about the number of cells mismatch is the expected behavior. Your solution looks OK.

Okay, then I'll update the rest of bimaterial tests accordingly test/GridapEmbeddedTests/EmbeddedBimaterialPoissonCutFEMTests.jl and
examples/EmbeddedBimaterialPoissonCutFEMTests. If you agree, I'll push directly to the gridap:update_to_Gridap_v0.15.0.

[fverdugo]

Yes you can push directly. Thanks!