Inspiration from today's demo meetup: @Nilesh Trivedi casually tossed out the idea that there are 2 kinds of edges between software units: edge (visible line) and containment (invisible(?), but, easy to calculate using bounding boxes). I think that containment is important, but, I've been avoiding it because I could only think of complicated ways to represent it. If it's "just" an edge, then the problem is solved 🤦. I've used Prolog (exhaustive search using backtracking) in the past to infer edge information from JSON (SWIPL even knows how to inhale and exhale JSON). Now that backtracking is no longer a verbotten technique (OhmJS, Prolog, miniKanren, Nils Holm's Scheme code & transpilations of that code to other PLs, cl-holm, etc.) it should be straight-forward to create DPLs that use "containment" visual syntax. Hmmm...

Relevant for visual layout of such graphs: https://arxiv.org/abs/2408.04045