SNA

Here you can download the prerelease of the new sna module: http://sna.drupaler.net/?q=en/node/34.
It's possible to install and play with it but still lack features and bugfixes. I wait for the feedbacks.

At my blog you can track my work altough I think I should create a summary here too.
Let's see:

Currently my problem is to find a good algorithm to visualize a tree. The problem is:

• determine each point's coordinates - find a layout
• draw the layout

This tree is not weighted, directed one - breadth-first search output for example. The second point is very obvious. The first one is not easy, especially for very big trees (~3000 points)

Here are two finished graph visualization what represents two large community sites:

• www.hup.hu
• 2006.planet-soc.com

These are the newest results of my work.

I would like to thank to Trey (www.hup.hu) to give me one of the biggest Hungarian Drupal-based site's sql dump. I could test to create a graph from such a huge data (the whole sql dump is about 180 Mb) . I used serialize to store the graph. The graph has ~70 000 edges (weighted and directed) but the memory footprint (and the serialized file) is just ~770 Kb. As far as I know a gdbm-type file isn't very comfortable to store adjacentcy list while the Dijkstra algorithm has such type of input. The sna module is likely to use serialize to store the graph.

Finding out, how data is related to another one is easy to be represented as a "Model View Controller".

As an example: If relations between users are build through the buddy.module we know the direct relation between 2 users. What we want to figure out is the "shortest path" between 2 users, who are not buddied.

M: Array of known relations (optionally with weight)
V: Visualisation of the path like shown at http://sna.drupaler.net/?q=de/node/13
C: Dijkstra's algorithm for "Shortest Path"; http://sna.drupaler.net/?q=de/node/10 ; sna.module

If following relations-Model (with weights) exists:

An easy and effective way would be to connect users through the buddy.module ;)

But that is only one point of view and is not representive.

If you are thinking about what can be "connected", it would be:

• Users (user)
• Nodes (node)
• Categories (taxonomy)

e.g.:
User-User: Who knows who? (buddy, ...) / Who has a similar profile? (profile, ...)
User-Node: Who wrote what? (core) / Who reads what? (statistics, ...)
Node-Node: Which are similar? (taxonomy, ...)
User-Category: Who is expert for a theme?

If you are using social network service like orkut, openbc, MySpace or other:

It's exciting to read about the Social Networking Analysis (SNA) Tool that is 1 of 14 2006 Summer of Code projects selected to write Drupal code with the financial support of Google.

Kudos to Robert Douglass and Angela Byron for all their excellent work in helping put together the 2006 SOC!

The basic SNA module is an analysis tool that will map and analyze how nodes and users are interlinked. It is useful for knowledge management in organizations because it provides an insight on what nodes have greatest value and which users are frequently tapped by other users for knowledge. Since node analysis (eg most popular nodes) is already existent in other modules, we will focus here on an analysis of user networks.