One of the things we've wanted to do recently at Divmod is track development status of milestones as well as gain some perspective on repository history/trends in activity. The first thing I did? Looked at SQL for about 2 and a half minutes. The second thing I did? Fired up a python instance and imported storm.locals :-)
With a few minutes of typing and looking stuff up (e.g., how to define compound keys for an already extant schema/db), I was up and running and was able to concentrate fully on the problem at hand (reports) and how to represent data visually (matplotlib). Now that's how an ORM is supposed to work :-)
Here are the schemas I defined in python:
Note the 1-to-m relationship of revision to nodes changed -- so easy :-)
This is exactly what a (good) tool is supposed to let you do: focus on and solve the larger problem at hand, not get lost fixing tools. With the schemas defined, the queries took literal seconds to write, I had my data, and was able to start generating textual summaries as well as processing the queries for graphs.
Wanna see some graphs? Below is a slide show of some graphs (just a few of the hundreds that are generated) representing the work of the Divmod dev team over various points in time on a secondary trac instance (not our main, Divmod trac instance). Though I haven't included them in the slide show, the code also generates graphs per-user, per-time-period.
Created with Admarket's flickrSLiDR.
Update: Thanks Nafai for poking me about the oversight of not giving shouts out for the graphing tool: matplotlib (of which I have been a huge fan of since I first heard about it years ago, having suffered through proprietary and kludgey Matlab as a physics student).