Makie equivalent of matplotlib's `PolyCollection`

Specific Domains Visualization
1

I’m looking to reproduce something like the following in Makie:

using PyPlot

const PolyCollection = PyPlot.matplotlib.collections.PolyCollection
polygon = [[0.0, 0.0], [1.0, 0.0], [0.0, 1.0]]
# force new axes
ax = PyPlot.axes(label=string(Base.UUID(rand(UInt128))))
ax.set_aspect(1.0)

# quite clumbsy
grid(xs, ys) = collect(reinterpret(reshape, Int, reshape(collect(Iterators.product(xs, ys)), :))')

offsets = grid(1:10, 1:15)

# based off https://github.com/matplotlib/matplotlib/blob/40862a32e86ed3a9a67790dc249da13971f1a37d/lib/matplotlib/axes/_axes.py#L4886-L4892
poly_collection = PolyCollection(
    [polygon],
    offsets = offsets,
    edgecolors = "face",
    facecolors = rand(n, 4),
    transOffset = matplotlib.transforms.AffineDeltaTransform(ax.transData),
)

(xmin, xmax) = extrema(offsets[:, 1])
(ymin, ymax) = extrema(offsets[:, 2])
corners = ((xmin, ymin), (xmax, ymax))
ax.update_datalim(corners)
ax._request_autoscale_view(tight=true)

ax.add_collection(poly_collection, autolim=false)

display(ax.figure)
ax.figure.savefig("poly_collection.svg")

image

Extra bonus if the SVG output can use an href instead of inlining the polygon path repeatedly.

This is how matplotlib implements e.g. hexbin.

2

This would be one way although I can’t claim the SVG href bonus, that would require much more optimization in CairoMakie. In my approach you also have to shift each polygon where you want it as a whole, which is a bit more work (especially with rotations), but the example is still only a couple lines, so :man_shrugging:

using CairoMakie

pol = Point2f0[(0, 0), (0, 1), (1, 0)]

offsets = vec(Point2f0.(1:10, (1:15)'))

poly([pol .+ o for o in offsets],
    color = rand(RGBf0, length(offsets)),
    axis = (aspect = DataAspect(),))

grafik

2 Likes
3

Thanks! That’s certainly an option.

I need to update the colors of each of the polygons, and I think it would not work well if each were an Observable, instead of having one Observable array.

4

I’m not sure I understand. You do want to use an observable with an array of colors? That should already work with the method I showed, no?

5

Ah, yes, indeed!