The Eclipse Usage Data Collector has been running for several months now. I spent a few minutes yesterday cleaning up the presentation of the data.
The Usage Data Collector page on eclipse.org now has fresh links to the report data. You’ll notice some changes. For example, all the report data is available in CSV (comma-separated values) format files for download and manipulation using your favourite software. I have the data in two formats. The first format, labeled “csv” contains the data in straight tabular form:
As you can see, it’s broken down by year/month, view id, execute count, and user count. In the case of a view, “execute count” is the number of times that the view has been activated (i.e. brought into focus by the user). As implied by the name, “user count” refers to the number of users who have activated the view at least once in the month.
The “trend” download is provided as a convenience. It’s the same data, but with the execute count and user count stretched out vertically; each view id is represented once in the first column. Hopefully you get the idea.
There are similar files for bundle activations, command executions, editor activations, and perspective activations. All of these reports shown only org.eclipse.* bundles that are used by more than five people. It’s relatively easy to add other types of reports, so if something is missing, please ask.
Since my favourite software is Eclipse, I used BIRT to wander through the data a bit and developed a couple of charts like this one which compares the number of times the various views in org.eclipse.ui.views are activated:
This chart is telling me a couple of things:
- March isn’t over yet
- A very large proportion of our community doesn’t use the TaskList, PropertySheet, or ProgressView
- EditorList and Bookmark view are almost never used (I’m not even sure how to find EditorList)
- A heck of a lot of people take vacations in August
This data makes me worry a little about the ResourceNavigator death march. There still seems to be a lot of folks using this view, so we need to be careful with the transition plan. It’s very good that Francis is being very public about the process (though, Francis—if you’re reading this—a blog posting or two would broaden audience).
My next step will be to fully automate the generation of these reports. Currently, they are mostly generated automatically. I have on further step. FWIW, when I do implement the full automation, the URLs of the CSV files will change.
