I'm prototyping some code that exposes an edge_list iterator interface to a database cursor. I've defined a forward_iterator using iterator_facade and can use that to create an edge_list graph. I'm running into a problem however calling bellman_ford_shortest_paths on the graph. I get boost_bellman_ford.cpp:18: error: no matching function for call to 'bellman_ford_shortest_paths(Graph&, position_t&, edge_weight_map, long long int*, double*)' The code leading up to that looks like: position_t N = getNumVertices(); edge_iterator iter; DEBUG_MESSAGE("Contructed iterator"); Graph g(iter, iter.end()); DEBUG_MESSAGE("Constructed edge list graph"); std::vector<double> distance(N, std::numeric_limits<double>::max()); distance[0] = 0.0; std::vector<position_t> parent(N); for ( position_t i = 0; i < N; ++i ) parent[i] = i; boost::bellman_ford_shortest_paths(g, N, edge_weight_map(), &parent[0], &distance[0]); The edge iterator is linked to the database cursor on construction. The problem I am guessing is in how I define the edge_weight_map. It looks like struct edge_weight_map { typedef cost_t value_type; typedef cost_t * reference; typedef edge_t key_type; typedef boost::readable_property_map_tag category; }; boost::property_traits<edge_weight_map>::value_type get(edge_weight_map map, boost::property_traits<edge_weight_map>::key_type key) { return key.cost; } It makes no sense for me to repeatedly query the database for the edge weight, so I attach it as a third field in the edge_descriptor. The get function merely returns the third component of the edge struct. I'm sort of assuming the way I am defining the edge_weight_map is the problem. Any insights? Also, I noticed the simply constructing the edge_list graph traverses the entire iterator range. Is there some copying going on? Can I override this behavior? THK -- Timothy H. Keitt http://www.keittlab.org/