On Fri, 8 Nov 2013, Sensei wrote:
On Friday, November 8, 2013, Jeremiah Willcock wrote: You can pass in a sequence (iterator range) of edges and the compressed_sparse_row_graph code will automatically count and sort them. Can you get that as a single entity (or a way to compute the edges), at least?
Well, yes. Edges are computationally intensive, but can be computed. I'd like to avoid multiple copies of nodes/edges, since the graph can easily occupy 10 gb of ram.
OK.
I have the list of nodes in a std container, for completeness of information. The edges are calculated by a functional on node pairs (ie, I need n^2 operations).
Even to find just n*k edges?
I could create a list of pairs of nodes for edges, but I think it would defeat the purpose of using a graph, at least, if creating the graph involves copying all edges. If there were a sort of "moving semantics" with this respect, it would help a lot.
If you can get aligned vectors of edge sources, targets, and edge properties (if you have them), there is a constructor that effectively moves out of those. -- Jeremiah Willcock