​To generalize the method, I think we need:

* a starting node in the graph, which will be the root of the tree

* a rule to find children of a node; a generic rule could be:

for a node of type T, the children are found by following predicates (P1, P2, ... PN)​ and selecting objects node following conditions (C1, C2...CN)

other nodes linked to a node of the tree are either ignored or taken as 'value' of the node

 

What do you think of that generic approach?

--
Jean-Claude Moissinac