Hi Mira:

here is the way I thought we could define a useful family 
of derived edges by writing strategies instead of writing
Java code:

Using your example:

define -> v$Vertex,d,a$Adjacency (Graph g)
 {-> g$Graph  Adjacency stop
  -> a$Adjacency v$Vertex through -> *,source,*}
{precondition: for all v there is exactly one a such that:
 -> a$Adjacency v$Vertex through -> *,source,*
}

This derived edge could be added just for the purpose of the DFT
computations.

An alternative would be to actually store the link instead
of recomputing it each time, using a link visitor:

LinkVisitor(Strategy 
  {-> Graph  Adjacency stop
   -> a$Adjacency Vertex through -> *,source,*}) {
   when-traversing Vertex
     Adjacency d;
     (@ d = a; @)
}
{precondition: for all v there is exactly one a such that:
 -> a$Adjacency v$Vertex through -> *,source,*

Both solutions are longer than your back-edge solution (<- source)
but maybe also easier to implement. The idea is in both cases the same:
Use strategies to define "object-linking code" instead of using
oo code.

-- Karl