User agents would be great to get a hold of in an easy/aggregated format :). I'm planning on redoing my gadgetry script (which runs over the analytics slaves inside the cluster, rather than the TS), so if people have specific questions - last edits to gadgets, that sort of thing - drop me a note.

On 28 September 2013 04:55, Maarten Dammers <> wrote:
Hi Sumana,
(in three months I guess)

Op 4-7-2013 0:22, Sumana Harihareswara schreef:

Summary: we have some new stats regarding gadget usage across WMF sites,
but I'd like more analysis of gadget & bot usage.

Oliver Keyes has some code and results up at to
analyze "data around gadgets being used on various wikimedia projects":
A while ago I did some some simple analytics on user language settings. It's at
This was based on the user_properties table ( I'm a pretty sure you can grab from that table what gadgets are enabled. Should be straightforward to write a script that goes over all wiki's, grabs these settings and produce a report. You need elevated access for this because this is not normally exposed in the data we have at the Toolserver or Toollabs.

My team would also like to know:
* who maintains the most popular gadgets? (so we can invite them to
hackathons, help get them training, get those gadgets localised and
ported to other wikis, and so on)
All gadgets have a prefix, it should be straightforward to write a script to see who do the most edits on these.

* when were the gadgets last updated? (so we can identify stale ones
that enthusiastic volunteers could take over maintaining)
Easy one too.

* similar stats regarding bot usage -- what bots are making the most
edits, or edits that in aggregate change the most bytes? who owns those
bots? what wikis are they active on? (so we can help maintainers better,
ensure they hear about API breaking changes, etc., and develop a bot
inventory/directory to make it easier for other wikis' users to start
using useful bots)
Erik already pointed out some useful statistics on that. As you might know I'm one of the maintainers of Pywikipedia. In the framework we send the user-agent in this format: USER_AGENT_FORMAT = '{script}/r{version[rev]} Pywikipediabot/1.0' or '{script}/r{version[rev]} Pywikipediabot/2.0'. We really don't know who is running what. Maybe you can gather some data (maybe with more samples) to answer to following questions:
* Usage per script
* What version are people running? (we switched from svn to git some time ago, but a lot of people seem to be still using svn)
 * Total number of distinct users using this (distinct ipaddresses)
* The 1.0 / 2.0 ratio


Analytics mailing list

Oliver Keyes
Product Analyst
Wikimedia Foundation