Unidirectional != undirected, which is the current intended meaning.
Example: linear graph
o -> o ->o
this graph is directed and unidirectional;
> o <> o
this graph is undirected and bi-directional – assuming edge weights == 1
In the current implementation the attribute unidirectional= 0 for the default demo dataset means that the matrix is directed...
I guess this attribute is used by the BCT algorithms. In that case two attributes are needed:
so to have the 4 types of graphs that can be handled: weighted-directed, weighted-undirected, binary-directed, binary-undirected.