I’ve updated my twitter client, undulatus. It no longer uses sqlite via sqlalchemy, but instead shoves all the tweets into a CouchDB database. This fixes an annoying problem with threads and locking, and also makes it much easier to do analysis on tweets.

I’ve also put all the deps into git submodules, so it’s now pretty easy to get the program running.

Now that I’ve played with it a bit, I have to say CouchDB is pretty neat. I can see why the hype – it is a better way of dealing with certain problems. I did consider using MongoDB (which seems to be pulling in more oxygen as a project) but it doesn’t run on ARM – and I want to keep this available to people running it on little ARM linux devices.

As part of this work I migrated my old tweet database, and discovered I have over three million tweets. They mostly go back to March last year – I had more, but lost them in an accidental database corruption. I’m now running CouchDB on my laptop and replicating from my server – kind of neat :-)