How to evaluate FEFunction on cartesian grid

[haasal]

I have a set of solution vectors for different Eigenmodes and want to export these solutions as 2D matrices.
Currently I'm doing

uh = FEFunction(system.Ug, sol.vectors[:, mode])

function (uh::FEFunction)(x::T, y::T; mode::Integer=1) where T<:Real
    pt = Gridap.Point([x, y])

    try
        uh(pt)
    catch e
        if isa(e, AssertionError) # Outside bounds
            0
        else
            print(e)
        end
    end
end

But this is really slow and I feel like there has to be a more elegant solution

[JordiManyer]

https://gridap.github.io/Tutorials/dev/pages/t015_interpolation_fe/

[haasal]

I read that but this didn't solve the issue that I had to check the bounds using try/catch on every point.

Edit: Ah nvm. The solution is probably interpolating onto a new (quadratic) triangulation, right?

[JordiManyer]

If you need the try/catch is because some of your points are outside the domain, i.e the tree-search algorithm cannot find a cell containing some of your points.

[JordiManyer]

However this will not solve your problem, but there is nothing you can do. Evaluating something that is defined piecewise on cells on generic points is just inefficient due to the fact that the code has to find the cell to which the point belongs to. If you want something efficient, you have to classify your points into cells first.

[haasal]

I'm sorry but I have no idea how to do that. Can you please point me into some direction? I.e. functions to use

[JordiManyer]

You can use Gridap.CellData.compute_cell_points_from_vector_of_points(xs::AbstractVector{<:Point}, trian::Triangulation, domain_style::PhysicalDomain).