it is (obviously) correct that hypergraphs are "undirected", since those structures mentioned below are "directed hypergraphs".
Since an incidence matrix can be used to implement either, it doesn't make sense to (to me) discuss them exclusively. The fact that the term "hypergraph" denotes to its "undirected" property does not exclude the data structure from being used to implement both. This is directly analogous to the definitions of the adjacency list and matrix in the BGL. Hypergraphs themselves are already such complicated beasts, that I would
predict that no useful library can come out of a concept combining "directed hypergraphs" with hypergraphs.
I don't think that "It's challenging, therefore it's not useful" strikes the right tone here. Andrew Sutton andrew.n.sutton@gmail.com