And added 2nd parameter for "vtxt()" OpenSCAD module, allowing to not only label vertices with their vertex number.
Finally I added new option "-10" that does determine "vtype" for each vertex when selected (sides 1 and 2, and 0 indicating vertices on separating circle):
(this big diff is best viewed with github split mode)
https://github.com/Hermann-SW/planar_gr ... cf0decc0c8
Code: Select all
./rjs node.tetra.js graphs/C60.a white.70 -10
The reason for cleanup work was clear.
Adding new "vtype" related option is preparation for next step.
I had no success with single vertex change spring embedders.
Success happened when solving system of linear equations for determining vertex coordinates.
After vertex coordinates have been determined, for vertices of vtype 1 and 2, I will call "tutte.convex_face_coordinates()" again, but this time I will not pass the vertices on separating circle as fixed coordinates, but the vertices of vtype 1 and 2. That will determine new coordinates for vtype=0 vertices (on separating circle), by calling "tutte.convex_face_coordinates()" once. Inspecting the difference then will show whether repeated determination of vtype=1 and vtype=2 vertices, and then of vtype=0 vertices might approach "better" and "better" spring embeddings or not (whatever "better" will be) ...
P.S:
"better" might be more "regular" spring embedding for C60 like football:
