Subscribe to RSS feed

splitbrain.org - electronic brain surgery since 2001

Graphing DokuWiki (help needed)

I recently had the task to restructure a certain part of a wiki. What I needed was a quick overview on what pages and images are in that namespace and how they are linked currently. But there's no simple tool to do this.

Inspired by an old forum thread I first looked for a way to represent the link structure in a mindmap. But mindmaps are strictly hierarchical, but links in a wiki are not. The right term is probably bidirectional graph.

Well, the Unix standard tool to draw graphs is Graphviz and so I wrote a tool to gather the data I was interested in (pages and media files) and to create a graph description in the dot language. The result is pretty nice for relatively small namespaces:

Graphviz example

However if I'd wanted to use that on a whole wiki, it wouldn't be very useful. I needed something more interactive. That's when I found Gephi.

Gephi is a pretty powerful tool which seems to be able to do all kinds of stuff with graphs. It uses its own XML based format called GEXF which I quickly implemented for my tool. Standard results look quite okay already:

Gephi example

Of course I had to run the graphing tool on the whole dokuwiki.org wiki. The resulting 1.6 MB XML looks quite impressive in Gephi and I'm sure there is a ton of valuable information hidden in that graph. But I'm not experienced enough in Gephi or data analysis in general to unveil it.

Part of the whole dokuwiki.org graph

This is where you can help

Download the file below, fire up Gephi and create some stunning visualizations, in depth analysises or whatever else you can find. Please share your creations with us in the comments.

BTW. the GEXF format allows to attach additional data to the nodes and I added file sizes and timestamps which could maybe used for visualization.

Tags:
graph, visualization, dokuwiki, gephi, graphviz
Similar posts:
Posted on Monday, August the 2nd 2010 (4 years ago).

Comments?

blog comments powered by Disqus