`tangents` and `waypoints` support for self-loops about graphmakie.jl HOT 2 CLOSED

The way tangets work for "normal" edges is not directly extendable to self edges. The tanget is only the directional information. The placement of bezier helper points depends on the base distance between src and dst vertices. This makes sense because you can achive a similar look regardless of distance of the vertices.

g = SimpleGraph(6)
add_edge!(g, 1, 2); add_edge!(g, 3, 4);add_edge!(g, 5, 6)

fig, ax, p = graphplot(g; layout=_->([0,0],[1,0],[0,1],[2,1],[0,2],[3,2]),
                       tangents = (Point2(1,1), Point2(1,-1)))
ax.aspect = DataAspect()
GraphMakie.plot_controlpoints!(ax, p)

For self edges, where src and dst are the same position, all 4 positions will be identical and there is no loop.

So what you need there is some kind of absolute information in data space, the problem is not solvable purely with tangets and tfactor. Enter: selfedge_size, selfedge_width and selfedge_direction. The direction is not the outgoing direction, it is responsible for the direction of the center of the loop. The width is given in radians and tell the algorithm the angle between outgoing and incoming arc (one is width/2 left of direction, one is width/2 right of direction). The size just pushes the bezierpoints further out until the arc crosses direction at that distance from the node.

using GraphMakie.NetworkLayout: SquareGrid
g = SimpleGraph(4)
for i in 1:nv(g)
    add_edge!(g, i, i)
end
fig, ax, p = graphplot(g; layout=SquareGrid(),
                       selfedge_direction = Point2[(0,1), (1,1), (-1,-1), (0, -1)],
                       selfedge_width = [pi/5, pi/4, pi/3, pi/2],
                       selfedge_size = [1, 0.75, 0.5, 0.25])
ax.aspect = DataAspect()
GraphMakie.plot_controlpoints!(ax, p)

I think this is the minimum set of required parameters to specify whatever selfloop you want as long as it is symmetric arround direction and represantable as single cubic bezier spline. Not sure what what else to achieve with tangets.

Waypoints is a different story though, that should totally work. Should be an easy fix though: the selfedge branch should be only hif if there is no waypoint data for the selfedge.

from graphmakie.jl.

@hexaeder Thank you for your explanation! I can add a PR to support waypoints on self loops if that is something you're ok with adding. Let me know

from graphmakie.jl.