Roland SchmidtSep 6, 2024
Teun van den Brand
I'm just going to post this without any further context.
#rstats #ggplot2
Sep 5, 2024 at 4:26pmWeb
Show threadMichael SumnerSep 5, 2024

@teunbrand what code does that?

I had the beginnings in rgl::wire3d(quadmesh::quadmesh(z)) but it's not aligned to current rgl and new stuff is scattered in gh projects 🙏 (I was obsessed with quad- vs tri- meshes and the details of centre-based vs area-based rasters, and bilinear interp in {raster} which was borko)

Show threadTeun van den BrandSep 6, 2024
@mdsumner Oh this doesn't do anything fancy, it assure you it is quite low-tech. Mostly it is just `trans3d()` + custom code to make a transformation matrix (similar to the return value of `persp()`).
Show threadMichael SumnerSep 6, 2024
@teunbrand I'm impressed that it represents the wiremesh faithfully (of the rotated sinc z matrix in ?persp, I assume) these details rock my world
Show threadTeun van den BrandSep 6, 2024
@mdsumner The wiremesh effect was just converting the matrix to long format and using two `geom_path()`s: once with `group = x` and once with `group = y`. The hard bit was getting ggplot2 to adhere to a z-scale.
Show threadMichael SumnerSep 6, 2024
@teunbrand ah! Interesting
Show threadLorenzo IsellaSep 6, 2024
@teunbrand Are you bringing 3D plots to ggplot2?
Show threadTeun van den BrandSep 6, 2024
@larry77 All I'm saying at this point is that I might eliminate some barriers that currently prevent this.