Paul McMillan has been working hard at his dbgraphnav Google Summer of Code project, and it's now pretty much ready to use. I've installed it on TreeTapper (see an example here). This allows navigation of author and reference relationships, a visual complement to the coauthor and reference tables already present. It's a general tool allowing navigation of relational databases with a lot of user configuration options. After all Paul's work, and our frequent IRC conversations, it's good to see it running. One nice aspect of it is caching: GraphViz takes a long time to draw graphs, and there's a user-set option to cache by time or by using diff to only update graphs that have changed (the latter works remarkably fast). I'll be firing off a few scripts tonight to generate cached files for all the people and references in the database to make it easier for people to use.
One aspect of the project I've been disappointed by is GraphViz itself. When I proposed this project, it seemed like the logical choice to use, but I've been amazed at some of its limitations. For example, on graphs with many nodes, using "neato" layout (an undirected graph layout), some nodes are far from the rest of the graph (see Michael Donoghue's TreeTapper page for an example). This results in nodes in the center being tightly clumped while nodes at the edges have far too much whitespace around them. No matter what options we've tried for desired edge length, spring parameter, etc., we just can't make GraphViz pull those far off nodes in. Any suggestions? Other than that sort of annoyance, though, this should be a very useful addition to TreeTapper and perhaps other websites as well.