Hi Jason, On Sun, Aug 2, 2009 at 3:38 PM, Jason Grout<[email protected]> wrote: > > Minh Nguyen wrote: > >> Here I speak from my experience in trying to (unsuccessfully) update >> the NetworkX spkg to version 0.99. Currently, Sage is shipped with >> NetworkX 0.36 and I have created ticket #6041 >> >> http://trac.sagemath.org/sage_trac/ticket/6041 >> >> in order to get NetworkX 0.99 into the standard packages repository. >> Note that NetworkX 0.99 has been around over a year now, and version >> 1.0 is still in development for about the same amount of time. In this >> case, we would not be updating NetworkX that often, probably about >> once or twice a year on average. Where the release dates between two >> successive stable releases are greater than 3 months apart, or about a >> year apart, one would expect drastic changes in the API. This is what >> happened with the changes between NetworkX 0.36 and 0.99. The changes >> between these two versions break API compatibility between the two >> versions, and to the credit of the NetworkX developers these changes >> are well-documented. But with drastic changes, it can be difficult to >> upgrade an spkg to the newer version. Of course, here "difficult" is >> understood to mean someone like myself with little experience in >> updating spkg's, as opposed to reviewing updated spkg's. So on the one >> hand, I'm rather peeved that my attempt to upgrade to NetworkX 0.99 >> has been unsuccessful. But on the other hand, it has stimulated me to >> spend more time debugging the issue. > > > I thought part of the problem in this case was that extensive work > needed to be done to upgrade our graph theory library to use the new > networkx functions, check all of the times we call it, etc. Some of > this is probably a lot easier now that Robert Miller has split off the > backends into self-contained classes, but we still call Networkx a lot; > doing a major upgrade of networkx seems like it would be a nontrivial > amount of work. In particular, they redid their design (e.g., splitting > out labeled graphs from other types of graphs, etc.), if I recall > correctly, which might affect the entire structure of our graph theory > functionality. > > But I haven't tried upgrading, so I'm just speaking from my impressions. > I have looked at the new networkx, though. It looks nice.
Before Robert Miller's extensive work on c_graph, I spent a few days attempting to do the upgrading to NetworkX 0.99. It turned out that the newer version broke a lot of doctests. So I defered the upgrading until Robert's c_graph patches are merged, which they have been. That way we would not be stepping over each other's toes. Now that the c_graph code are in, I need to find more time to try again. -- Regards Minh Van Nguyen --~--~---------~--~----~------------~-------~--~----~ To post to this group, send an email to [email protected] To unsubscribe from this group, send an email to [email protected] For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
