Layouts of graphs with multiple components

[thchr]

Several of the layout algorithms produce not-very-useful layouts if the input graph is not connected, i.e., has more than one connected component.

For example:

using GLMakie, GraphMakie, Graphs
g = wheel_graph(4)
es = collect(edges(g))
g2 = Graph(vcat(es, [Edge(e.src + nv(g), e.dst + nv(g)) for e in es])) # make two unconnected copies of `g`
graphplot(g2; layout=Spring())

Produces:

Which has at least two issues:

1. The two graphs are seemingly unnecessarily far apart.
2. The layout of each individual component is worse than it would be for a single component (compare with graphplot(g, layout=Spring()) below):

   

From testing, Stress, Spectral, and especially SFDP have this issue. Stress probably also has the issue, but it is less pronounced. E.g., here's graphplot(g2; layout=SFDP()):

I think it would make sense to apply these layout algorithms on the connected components of a graph individually, and then somehow layout their components subsequently.

[hexaeder]

Ah yes, that recently came up in a GraphMakie issue, for Stress I added a workaround.

Instead of handling this in each and every layout, an easy way might be to define a new meta layout algorithm:

layout = ComponentwiseLayout(sublayout=Spring())

(bad name) which splits the graph in components, applies the sublayout per component and uses some layout algorithm to place the results of the individual sub layouts.

If you're willing to work on that i'd be happy to take a PR and help along the way if necessary :)