Understanding Meshes and GeoTables visualisation capabilities

[juliohm]

@BambOoxX

My question is : is it possible to (or more likely how hard is it) to add application-specific element types for storage in a GeoTable and for visualization with viz

You can add a new geometry in Meshes.jl in a PR, and it will automatically work with the GeoTable, viz, etc.

@BambOoxX

As an example, I just may want to plot arrows or pyramids based on a geotable with specific elements. Question is, how can one write the proper recipe for that, should I just add some Makie based recipes, or does it make sense to rely on viz or else…

You can already do that by creating Ray and Pyramid geometries for example. You can place all these geometries into a GeometrySet and that will be a valid Domain. The Mesh is a specific subtype of Domain that is described in terms of vertices, connectivities, etc.

julia> using Meshes

julia> r = rand(Ray{3,Float64})
Ray{3,Float64}
├─ p: Point(0.8544621114647082, 0.007290166206372128, 0.8953959324059896)
└─ v: Vec(0.9568652328665942, 0.611726781018889, 0.27683459777111596)

julia> p = rand(Pyramid{3,Float64})
Pyramid{3,Float64}
├─ Point(0.5245605139763986, 0.31180747839374345, 0.8006794534270958)
├─ Point(0.17999965079093627, 0.2960387752699939, 0.23121858402483375)
├─ Point(0.16164742029403556, 0.5896417963031546, 0.0005634795956463989)
├─ Point(0.5200829045516021, 0.23759556572538754, 0.36635011529535155)
└─ Point(0.8140821418410266, 0.522107532502354, 0.5031798558753174)

julia> d = GeometrySet([r, p])
2 GeometrySet{3,Float64}
├─ Ray(p: (0.854462, 0.00729017, 0.895396), v: (0.956865, 0.611727, 0.276835))
└─ Pyramid((0.524561, 0.311807, 0.800679), ..., (0.814082, 0.522108, 0.50318))

julia> import GLMakie as Mke

julia> viz(rand(Ray{3,Float64}, 100))

@BambOoxX

FYI, my environment is completely up to date, but the points plotted with viz look transparent. I thought that could be just a z-level problem/artifact during the plotting, but the additional scatter does not suffer from that it seems. Regarding the markers themselves, I was wondering why scatter was preferred evend for 3D data plots, instead of meshscatter.

Here is the full recipe to plot a PointSet, which is constructed when you ask to viz the vertices of the mesh:

JuliaGeometry/Meshes.jl/blob/4a0d73fb7ba6f010f4e55bad152e0a61166302b2/ext/geometryset.jl#L68-L84

We simply call Makie.scatter with the overdraw=true keyword to address issues with large point sets, see Hide stroke in 3D scatter using Makie.jl.

@BambOoxX

is it possible to define coordinate transformations on the domain based on information in the table of a geotable ? As far as I understood from the book, you propose coordinate transforms, that modify the domain with scalar transforms (1 for the whole mesh e.g. rotation or translation), but no coordinate transforms that can use the information in the table, do you ?

The CoordinateTransform is a special case of GeometricTransform that simply applies the same function to each point of the domain individually. These transforms are defined in terms of parameters, they do not live in the columns of the GeoTable. If you want to design a transform that takes both the values and the domain of the GeoTable, then you are after a general TableTransform. See the hierarchy of transforms here: Transforms · GeoStats.jl

I think the easiest way forward is to start adding missing geometries in Meshes.jl, then trying to visualize them. After that, we can come up with higher-level transforms that meet your requirements, and a find a good home for them. Meshes.jl only defines GeometricTransform that act on the domain only, they don’t have access to the table of features.