BGL isomorphism vertex invariant interface change

[Evan Driscoll <driscoll@…>]

(This is the last isomorphism ticket I think. :-))

I have one more suggestion about how to handle vertex invariants. To determine whether u and v should be considered equivalent by the isomorphism, instead of providing two functors f and g, calling f(u) and g(v) to get a pair of integers, and finally comparing the result, I think it would be cleaner to have a "binary" predicate p which would be called as p(u,v). (In reality of course you'd probably want to have it receive g1 and g2 as well, so it'd really be a 4-arg function.) This should also let you get rid of the vertex_max_invariant parameter perhaps.

I'm not sure of how this would impact the performance or how you'd do sorting, but you could probably do it by doing an N^{2 traversal and recording for each u in g1 how many vs in g2 are considered equivalent. This is a bit different from what you're doing now, I think.}

(The benefit of this would be it becomes easier to impose extra conditions on the isomorphism. Instead of figuring out how to encode everything into an integer -- and worrying about the fact that you make an array of size vertex_max_invariant, so you can't even use all that much of that size -- you can just test it directly.)