Hello Jeremy, I think the parallel edges are dealt just fine. I have thought it's impossible to be "two edges, one in the subgraph and one not in the subgraph" under the induced subgraph approach. Do you mean the second edge can be interpreted as not in the subgraph? Thanks, Hideaki Hiraki At Thu, 25 Mar 2004 11:29:52 +1100, Jeremy Siek wrote:
Hello Hiraki,
I don't think that the "induced" subgraph approach taken by boost::subgraph is the problem. There is no problem with interpreting the below graph and subgraph as an induced subgraph. There just happen to be two edges, one in the subgraph and one not in the subgraph.
I think instead the problem is a bug in the read_graphviz function with respect to dealing with parallel edges (two edges with the same source and target). I'll look into fixing the problem.
Cheers, Jeremy
On Mar 16, 2004, at 3:47 AM, HIRAKI Hideaki wrote:
Q3) Why boost::subgraph isn't general subgraph but "induced" subgraph? BGL seems to have a fundamental restriction on reading graphviz files. The graph data read in read_graphviz() is stored in the boost::subgraph<> where any edge connecting vertices of a subgraph is assumed to be part of the subgraph. This is against the semantics of the graphviz format. For example, digraph g0 { edge [color=blue] subgraph g1 { edge [color=red] n1 -> n2 } n1 -> n2 } this graph has one edge in the subgraph g1 (red) and the other not in g1 (blue). But the graph read by BGL must have both edges belonging to g1.