I am attempting to plot a function that has a discontinuity, consider for example,
function f(x, y)
if x^2 + y^2 ≤ 1/2
return 1.
else
return 0.
end
end
If I use the surface function from Plots.jl I obtain the following plot
surface(-1:0.01:1, -1:0.01:1, f)
Is there a way to plot the function without surface automatically interpolating the discontinuity points?
I don’t know of an automatic method, but in general, to obtain the same result, NaNs are inserted around the discontinuities to prevent interpolation:
using Plots; plotlyjs()
function f(x, y, δ)
r = x^2 + y^2
if r ≤ 1/2 - δ
return 1.0
elseif r ≥ 1/2 + δ
return 0.0
end
return NaN
end
δ = 0.01
x = y = -1.0:δ:1.0
Plots.surface(x, y, (x,y)->f(x,y,δ))
# QC mask of NaN:
heatmap(x, y, (x,y)->f(x,y,δ), ratio=1, lims=(-1,1))
Thank you for your answer!
Yeah, I thought about this, but in my case the discontinuity points are hard to compute. They are determined by an implicit equation, which I would have to solve with a root finder. I can do that, but it is not that elegant.
Ok, check if this automatic method could work in your case:
using Plots; plotlyjs(size=(800,600), dpi=600)
f(x, y) = x^2 + y^2/4 ≤ 1/2 ? 1. : 0.
x = -1.0:0.01:1.0
y = -2.0:0.01:2.0
z = f.(x,y')
ϵ = 0.5
∇x = vcat(diff(z, dims=1), zeros(1, length(y)))
∇y = hcat(diff(z, dims=2), zeros(length(x)))
z[hypot.(∇x, ∇y) .> ϵ] .= NaN
surface(x, y, z', xlabel="x", ylabel="y", zlabel="z")
NB: changed the circular disc to an ellipse for easier debugging
@jheinen, Plots.jl with gr() backend seems to exhibit several bugs on this example.
Bug on aspect_ratio=:equal, which is not applied to the z-axes:
surface(x, y, z', ratio=:equal)
Bug on the color scale limits if we set zlims to fix the aspect_ratio bug above:
surface(x, y, z', ratio=:equal, zlims=(-2,2))