Recent Projects: Groovebot

This is the first of a number of posts meant to overview the things I've been doing recently. First up on the list is the long-overdue Groovebot update.

Before this summer, Groovebot 'worked' on Spotify, and also had a fairly useful but under-tested MPD backend that worked for me when I could get MPD running properly. Unfortunately, I no longer had an active Spotify subscription, which meant that I couldn't use the Spotify backend, and I had moved off of RIT's campus, meaning I no longer had a ~100Mb connection to my fileserver, so streaming files that way was more problematic.

While looking around for other things I could use to get Groovebot running on, I rediscovered Pithos, a Python frontend to Pandora. Using Pithos as an authenticator, I could get URLs to music files that I could play if I could figure out how to play them in Python. As a temporary test, I just sent them to mplayer to play the files. This worked well enough, but had the same problem I had encountered with Spotify, namely that I could not get a callback properly positioned to fire when the song finished.

I put the project aside for a time, until Justin pointed out GStreamer is made to do this sort of thing. With this in mind, I took another look at the task, first trying to run GStreamer in its own event loop on top of Twisted, then giving up and processing events on a tick. This means I have to respond to every event GStreamer throws out (which is a fair number of them), but most of them can be thrown out without looking at them. Some more tweaking later, and PandBot was born, able to connect to a user's Pandora station, play music, and thumb songs up or down based on user response in the IRC channel.

At this point, I had been writing new bots each time I added a new backend, with a lot of repeated code in each, leading to subtle bugs and API implementations that were slowly drifting apart based on individual needs. This made me a little annoyed, but I didn't have any particular need to fix it, so I left it alone. Until an accidental click around the Spotify website left me with another Premium subscription (and $10 poorer...) I poked around on SpotBot to see if it still worked. Turns out it did... just not on the DJ computer in the FOSSBox. Turns out that spytify, the python bindings for despotify have gotten a little stale in the interim, and no longer compile with modern versions of Cython. This meant I should probably break down and get an API key from Spotify and use the proper official API instead.

However, in working on MPD and Pandora bots in the meantime had given me a number of ideas and fixes that would be problematic to port back to Spotify, which meant I had finally gotten an excuse to try to merge the three separate codebases into one unified version. Starting with MPD and Pandora, I merged the files and ran the result through a three-way diff, trying to identify important parts of each bot to make sure each worked properly after the merge. Next, I looked into the architecture of pyspotify, and immediately ran into some trouble. pyspotify works differently to the backends I had been used to, having its own threading I would need to juggle alongside my own. Additionally, the code is not thread safe, which is a problem with the heavily-threaded Groovebot code.

So now, the status is that we are down to two bots. One is the ancient Grooveshark code that is in need of updating and integrating with the second bot. This second bot has all the work I've been doing lately and, some few remaining problems aside, mostly work. Ideally, I will soon have a better host for the bot that will let me run Spotify, but for now Pandora is working well enough.


Similar posts to this one

blogroll

social