An Experiment in Productivity
This post is about being productive. It is also about drawing charts, Google App Engine, and monkeys. But mostly about being productive.
For a couple years now I’ve been running a project night. Once a week, myself and a few friends get together to work on whatever projects we’re currently involved with. The goal is to sit down and get stuff done, in the company of other productive people.
Last week, a friend and I decided to collaborate on a service for recording and charting measurements. The basic problem we’d be solving: “I want to measure some property X every N seconds and then draw a chart of the data.” The service would take care of aggregating the data and drawing charts. The user of the service would just have to write a client to periodically report measurements.
We set ourselves a challenge: try to finish a (very) rough end-to-end prototype in 2.5 hours. It would have a simple API for reporting new data points, a web page where you could view the charts, and one example client. Here’s what we managed to get done:
The whole things is done using Django on Google App Engine. There’s a simple REST API for posting new measurements, and a simple web interface for viewing the charts (which are drawn using Google Chart API).
Some observations from the project:
Picking a name is hard. Unless you’re in a hurry, in which case it becomes easy: you just pick the first name someone shouts out that isn’t taken. Stats Monkey was the first name that passed this test for us.
Having two people sitting next to each other made it much easier to stay in scope. When one of us was about to waste time over-doing something or going off on a tangent, the other one would step in. I don’t think we would have finished the project if we hadn’t been sitting next to each other announcing what we were about to do.
We both already had a lot of experience with all the tools we were using. This really helped, because there was basically no setup time. Google Code and App Engine both have really low startup costs for a project like this. I think maybe only 10-15 minutes were spent getting a subversion repository and hosting.
We hit a lot of SVN conflicts. We also had trouble with incomplete syncs/commits because we were both used to p4 semantics (where a commit includes files from the entire workspace, not just the current directory). It would be worth trying to mitigate this if we try something like this again.
The code is full of bugs and security holes, and doesn’t have any tests. Since we just barely made our deadline, I don’t think this was a mistake. We didn’t hit any major roadblocks, but this was probably just luck.