GSOC-Task #1: Visualizing 2D and 3D Grid objects

[fverdugo]

cc @paurierap @ericneiva @amartinhuertas

This task is to be further defined in the kick-off meeting

Intro

Being able to visualize instances of Gridap.Geometry.Grid is the cornerstone to visualize other Gridap objects. So, it is a good place to start.

Given a 2D Grid object (3D and 1D will be covered in subsequent tasks), we typically want the following kind of visualizations (now they are done with paraview). Here, I am using this code to generate a simple grid of triangles:

using Gridap
using Gridap.ReferenceFEs
using Gridap.Geometry
using Gridap.Visualization

model = CartesianDiscreteModel((0.,1.5,0.,1.),(15,10)) |> simplexify
grid = get_grid(model)
celldata = rand(num_cells(grid))
nodaldata = rand(num_nodes(grid))
write_vtk_file(grid,"grid",celldata=["d1"=>celldata],nodaldata=["d2"=>nodaldata])

surface	surface and edges	prescribe edge width
set surface and edge color	set surface color via a cell field	set surface color via a nodal field
display colorbar	change colormap

TODO (outdated, see below for the updated steps)

• Familiarize with Makie.jl and find out which are the low level plotting command needed to reproduce these images
• Propose a user API to generate these figures in a convenient way
• Implement the API (possibly by reusing code in the src folder)
• Extend to quads

Notes

Most of the functionality we need is provided by the Makie function mesh, but it is likely that this does not cover all cases.

[fverdugo]

Steps:

• create a new branch gsoc2021-master (consensus branch, all test should pass)
• Create you own work branch e.g. pr-task1 (you are free to use this branch as you want)
• Clear code in GridapMakie module
• Add Gridap="0.16" in Project.toml [compat] section. You perhaps need to resolve your environment at this point. pkg> resolve.
• Implement conversion Griadp.Grid -> GeometryBasics.Mesh

function AbstractPlotting.convert_arguments(pt::PlotType, grid::Grid)
  convert_arguments(pt,UnstructuredGrid(grid))
end
function AbstractPlotting.convert_arguments(::PlotType, grid::UnstructuredGrid)  
  xs = get_node_coordinates(grid)
  Tp = eltype(eltype(xs))
  D = length(eltype(xs))
  ps = collect(reinterpret(GeometryBasics.Point{D,Tp},xs))
  cns = get_cell_node_ids(grid)
  Tf = eltype(eltype(cns))
  fs = collect(lazy_map(GeometryBasics.NgonFace{D+1,Tf},cns))
  m = GeometryBasics.Mesh(GeometryBasics.connect(ps,fs))
  m
end

• Try to reproduce the plots above using Makie functions like mesh and wireframe. Add this code into a test.
• Move from Travis to GH actions to run the tests.
• Explore the extension to quadrilateral elements (i.e., remove simplexify).

[amartinhuertas]

Some relevant links/info interchanged during 07/06/21 meeting:

Eric Neiva17:41
- run: sudo apt-get update && sudo apt-get install -y xorg-dev mesa-utils xvfb libgl1 freeglut3-dev libxrandr-dev libxinerama-dev libxcursor-dev libxi-dev libxext-dev
- run: DISPLAY=:0 xvfb-run -s '-screen 0 1024x768x24' julia --project=@. -e 'using Pkg; Pkg.test(coverage=true)'
Eric Neiva17:43
https://github.com/JuliaPlots/GLMakie.jl/blob/master/.github/workflows/ci.yml
https://github.com/JuliaPlots/GLMakie.jl/blob/3762f689ec785ff40b429c9ecca7d81a54895ed0/.github/workflows/ci.yml#L41
Eric Neiva17:45
https://github.com/JuliaPlots/GLMakie.jl#troubleshooting-opengl

[fverdugo]

@paurierap perhaps it is better to use convert_single_argument instead of convert_arguments.

[fverdugo]

Some updated steps after the meeting on 2021-06-11:

• gitignore .DS_STORE files
• Add dependences of tests in the library Project.toml via pkg manager (for the moment).
• import AbstractPlotting instead of Makie
• Try to define conversion for all plot types at once (if possible). Try to use PlotType PlotFunc.
• Extend conversion to quadrilaterals
  can makie represent quads?
  if yes: quad Grid -> quad Mesh -> Makie
  if no: quad Grid -> tri Mesh -> Makie (ok for plotting mesh, but not wireframe)

[paurierap]

Before the meeting on 21/06/2021, the conversions implemented from Gridap types to Makie types allow us to plot:

2D surface + set color (extended to quads too)	2D edges + set color and linewidth (only for simplices)	2D surface color via a nodal field + choice of colormap (and colorbar)
3D surface + set color (extended to quads too)	3D edges + set color and linewidth (only for simplices)	3D surface color via a nodal field + choice of colormap (and colorbar)

So far, we still need to address the drawing of quadrilateral edges in either dimension as well as the plotting of a cell field.

[fverdugo]

Code discussed in meeting 2021-06-21 for visualizing 3D meshes. (i.e., convert volume mesh to boundary mesh of triangles)

using Gridap
using Gridap.Geometry
using Gridap.Visualization


domain = (0,1,0,1,0,1)
cells = (2,2,2)
grid = CartesianGrid(domain,cells)

write_vtk_file(grid,"grid")

topo = GridTopology(grid)

labels = FaceLabeling(topo)

model = DiscreteModel(grid,topo,labels)

face_grid = Grid(ReferenceFE{2},model)


face_to_mask = get_face_mask(labels,"boundary",2)

@show num_cells(face_grid)
@show length(face_to_mask)

write_vtk_file(face_grid,"face_grid",celldata=["mask"=>collect(Int,face_to_mask)])

boundary_grid = GridPortion(face_grid,face_to_mask) |> simplexify

write_vtk_file(boundary_grid,"boundary_grid")

[fverdugo]

TODO steps

• CI integration i.e., copy file https://github.com/gridap/GridapPETSc.jl/blob/master/.github/workflows/ci.yml More info here: https://docs.github.com/en/actions/guides/about-continuous-integration
• For 3D conversions, use a boundary triangulation of triangles, see code above.
• Open a PR against gsoc2021-master branch

[fverdugo]

closed via #18