
Hi Jeremiah, The issue Tim brought up was regarding the case when the graph is not fully read in the main memory. Instead memory is allocated per out_edges invocation. For this to work the algorithm should not break range obtained from a out_edges call. However current implementation of dijkstra's does exactly that. -- an example of range split: "depth_first_search(g, vis, color, *vertices(g).first);" Tim proposed that he should be able to provide patch to fix such cases. I hope he has because for my current project I intend to process a graph stored externally. thanks Sandeep On Fri, Jun 26, 2009 at 3:22 PM, Jeremiah Willcock<jewillco@osl.iu.edu> wrote:
On Fri, 26 Jun 2009, Sandeep Gupta wrote:
Hi Tim/Andrew, Was wondering if the splitting of out_edges range got fixed.
Are you talking about minimizing the number of calls to vertices() and out_edges()? If so, that was put in a while ago. Please note the information from the previous emails that you should not rely on being able to create new containers for each call, though, since new algorithms might not respect this particular implementation detail.
-- Jeremiah Willcock _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost