Hello all. I have a project in involving enumeration of some combinatorial objects that makes use of canonical labeling of graphs to eliminate branches of a search tree that (up to automorphism) have already been handled at some prior stage in the search. I am finding canonical labelings is by far the most computationally expensive thing happening, so I'd like to try to speed it up by using nauty instead of the implementation that comes with sage.
I have a couple of questions about this: First, as I understand it there's an optional nauty spkg that, according to a dev list email thread I saw contains a python binding for nauty, here's a link for that:
On the other hand, having installed the nauty spkg and searched a bit, I don't see any way to access that binding from sage. So my first question is: Was this python binding included in the spkg and how does one access it if so?
The other question is about how much difference can be expected using nauty. Has anyone ever made a direct comparison between the two implementations of the algorithm?