please reblog my photoset etc
i've already made that joke
so i still don't have better graph grammars working but i realized that it's actually real easy to synthesize the necessary stitching geometry on polyhedra, so, i did that. theoretically i could do this for any rectified polyhedra, not just the mostly-hexagonal ones.
the real big issue is that i'm having a hard time thinking of any reasonable way to turn all those points into a coordinate space. flat, euclidean geometry is real easy to manage. nice simple distance metrics. easy ways to calculate relative angle and positioning of two arbitrary tiles. all of that goes away once you bring curvature into the picture. i was actually having similar problems trying to figure out geometry in the hyperbolic plane a while back, and it didn't really occur to me at the time that the fundamental problem was curvature, not specifically the hyperbolic curvature.
all of that is to say that to make a better graph embedding i'll need a robust polyhedra coordinate system, and that... seems difficult. i might try to patch up the graph code itself but it's not very useful at all without an embedder.
