Hi all, I am working on the failing doctest to upgrade our almost 2 year old NetworkX 1.2 to NetworkX 1.6 (see #12806, [1]) I have sorted out the doctest failures in graph.py and digraph.py; in order to solve the ones in generic_graph.py there is some design decision that needs to be made.
The clustering coefficient method in NetworkX used to have (for NX <= 1.4) different types of output regarding whether or not weights were used in the computation. In the older NX, clustering coefficients *without* weights were returned as a dictionary indexed by the graph vertices, whereas clustering coefficients *with* weights returned *a couple of dictionaries*, one of them as before (containing the clustering coefficients) and another one containing the weights. >From version 1.5 of NetworkX onwards, this behavior was unified so that the clustering function would only return the dictionary with the clustering coefficients, regardless weights are used or not. In my opinion this is a reasonable change (why would I want the weights if I asked for the clustering coefficients?) and we should change Sage to adopt this behavior. However, doing that requires breaking backwards compatibility with our clustering_coeff function for weighted graphs, so I am asking for a public discussion of this. As a separate issue, I think we should move some of the centrality measures from graph.py to generic_graph.py, as NetworkX can handle them for directed graphs just fine. Votes? Opinions? Cheers, J [1] http://trac.sagemath.org/sage_trac/ticket/12806 -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
