After a night of hacking with
Essentially we are requesting the latest posts feed every 10 seconds, parsing out the lj:mood tag and aggregating it up for 10 minutes, then we walk through and calculate the percentages of moods against a list of moods and values between 0 (good) to 100 (bad). This ends up with a plot on the scale somewhere between 0 and 100 with 50 in the middle. The list of moods was made very unscientifically by me and
Then we keep track of the min and max we have ever seen and the point we are currently in between, so we can self calibrated. We realised that using the absolute value wouldn't really work since it would never move very far away from midpoint.
All in 50 lines of code or so.
This morning it wasn't very happy.
But as the day progresses it has gotten a bit more happy.
Now I want to add lj:mood data to the atom stream.