On Wed, 29 Dec 2010, Ryan Lewis wrote:
I do mean adjacency list, and I agree this should be documented a lot better.
I'm not sure how the internals of adjacency_list<..> work, but I fundamentally don't understand why a subgraph can't be a method built on top of a simple adjacency list, it may be less efficient than something typed as a 'subgraph' capable graph, but I don't see why a graph shouldn't have one of it's fundamental operations work on it.
What would a built-in subgraph capability do? If you want it to produce an adjacency_list without copying the original graph, it would need to wrap all of the iteration functions (like filtered_graph does), leading to a slowdown for graphs whose subgraphs aren't used.
I will look into forcing our graph types to be subgraph < Graph > types. What is the right way to 'raise' this question to an appropriate developer?
Boost-users works for that. -- Jeremiah Willcock