weird interplay between scale and lightposition

[photor]

I was drawing a surface plot of some configuration of a quantum system, using the following code:

using Makie
l = w = 30.0
p = m = 120
x=l*[(i-0.5)/p-0.5 for i = 1:p] #x ranges in (-15,15)
y=w*[(i-0.5)/m-0.5 for i = 1:m] #y ranges in (-15,15)
scene = Scene()
limits = FRect3D((-1.5, -1.5, 0), (3.0, 3.0, 2.5))
update_limits!(scene, limits)
z = ... #data of the configuration stored in a (120,120) array z, ranging in (0, 2.25)
surf = surface!(scene, x/10, y/10, z, colorrange=(-150,50))[end]
wf = wireframe!(scene, x/10, y/10, lift(x->x, surf[3]), transparency=true, color=(:black, 0.1))

Here I divided x and y by 10 because the ranges of x and y are much larger than z, and I didn't know how to use the scale! method at that time. Then the output was the following image, with a very nice shining effect:

Later I knew the scale! method, so I revised the above code to the following one:

scene = Scene()
scale!(scene, (1,1,10))
limits = FRect3D((-15, -15, 0), (30, 30, 2.5))
update_limits!(scene, limits)
z = ... #data of the configuration stored in a (120,120) array z, ranging in (0, 2.25)
surf = surface!(scene, x, y, z, colorrange=(-150,50))[end]
wf = wireframe!(scene, x, y, lift(x->x, surf[3]), transparency=true, color=(:black, 0.1))

In this way, I can get the correct ticks on the x and y axes, but now the output becomes:

The shining effect is gone. I realized that this result may be caused by an improper lightposition parameter, and I saw from somewhere that the default lightposition is Vec3f0(1,2,3), so maybe I should set a lightposition=Vec3f0(10,20,3) in the latter code? However, that proved to have no help (actually made the case even worse). I also tried various kinds of lightposition parameters, without any success. Any solutions suggested?

[SimonDanisch]

Oh... that's likely because we don't scale the normals correctly...

[ffreyer]

You could try generating and plotting a mesh instead of the surface plot

using GeometryBasics
vertices = Point3f0.(x, y, z)[:]
N, M = size(x)
lin = LinearIndices(x)
faces = [QuadFace(
    lin[i, j], lin[i+1, j], lin[i+1, j+1], lin[i, j+1]
) for i in 1:N-1 for j in 1:M-1]
m = normal_mesh(vertices, faces)
mesh!(scene, m, color=z)

If that fixes it then this is might also be a problem caused by using uv coordinates here.

[photor]

You could try generating and plotting a mesh instead of the surface plot

using GeometryBasics
vertices = Point3f0.(x, y, z)[:]
N, M = size(x)
lin = LinearIndices(x)
faces = [QuadFace(
    lin[i, j], lin[i+1, j], lin[i+1, j+1], lin[i, j+1]
) for i in 1:N-1 for j in 1:M-1]
m = normal_mesh(vertices, faces)
mesh!(scene, m, color=z)

If that fixes it then this is might also be a problem caused by using uv coordinates here.

Thanks. Then I tried the follow code:

using Makie, GeometryBasics
function meshgrid(vx::AbstractVector{T}, vy::AbstractVector{T}) where {T}
vx'.+(a->0).(vy), (a->0).(vx').+vy
end
l = w = 30.0
p = m = 120
x=l*[(i-0.5)/p-0.5 for i = 1:p]
y=w*[(i-0.5)/m-0.5 for i = 1:m]
x,y=meshgrid(x,y)
z = ...
vertices = Point3f0.(x, y, z)[:]
N, M = size(z)
lin = LinearIndices(z)
faces = [QuadFace(
lin[i, j], lin[i+1, j], lin[i+1, j+1], lin[i, j+1]
) for i in 1:N-1 for j in 1:M-1]
scene=Scene()
scale!(scene, (1,1,10))
limits = FRect3D((-15, -15, 0), (30, 30, 2.5))
update_limits!(scene, limits)
surf = mesh!(scene, normal_mesh(vertices, faces), color = :orange)
wf = wireframe!(scene, x, y, z, transparency=true, color=(:black, 0.1))

and got the output:

which was not good. Even when I use the first method (dividing x and y by 10, without scale!) for mesh!, the case is no better:

Actually, when mesh! is used, it always seems that the light comes from below, and I can hardly change this situation by trying various kinds of lightposition parameters, weird. I also tried some simple functions, like cos(x^2+y^2), instead of my own data for mesh!, but the result is similar.

[photor]

BTW, is there a simple way to get a single color plot using surface! ? The keyword argument like color=:orange used in mesh! doesn't work for surface!

[ffreyer]

I think you can do colormap=(:green, :green).

The second mesh plot looks like the normals are inverted. If you do

faces = [QuadFace(
    lin[i, j], lin[i, j+1], lin[i+1, j+1], lin[i+1, j]
) for i in 1:N-1 for j in 1:M-1]

that should get fixed. But that doesn't solve your problem. I think Simon is right - there is something wrong with how GLMakie applies the scaling.

[ffreyer]

I think this is already be fixed on the SSAO branches (regardless of whether you use SSAO). You could try with JuliaPlots/AbstractPlotting.jl#424 and JuliaPlots/GLMakie.jl#97.

[photor]

I think you can do colormap=(:green, :green).

The second mesh plot looks like the normals are inverted. If you do

faces = [QuadFace(
    lin[i, j], lin[i, j+1], lin[i+1, j+1], lin[i+1, j]
) for i in 1:N-1 for j in 1:M-1]

that should get fixed. But that doesn't solve your problem. I think Simon is right - there is something wrong with how GLMakie applies the scaling.

Yes, you are right.

[ffreyer]

Looks fine to me now, both with scaling the plot and scaling the scene.