Wikipedia:WikiProject User warnings/Testing/Twinkle
Notes and coordination for testing of user warnings in Twinkle.
Background research
Most frequently used TW warnings
In order to figure out which templates to focus on in our A/B experiment with Twinkle templates, we first need to know which Twinkle templates are most used most frequently on user talk pages. To gather this sample, we queried the Enwiki database for a random sample of 1000 user talk page revisions made using Twinkle (identified by ([[WP:TW|TW]]) which is populated in Twinkle edit comments by default) since October 2010. Then we grabbed the whole list of user warning templates used by Twinkle and queried the API to get the text of each of these revisions. Then we pulled out all the template names from the HTML comments in the template messages and counted up the ones that appeared on the official list of Twinkle user warning templates.
Below is the list of the most-used Twinkle user warning templates from the last year. Should give us a good start to narrowing down which templates to update when we run the Twinkle study.
| twinkle user warning template | number of uses in random sample of 1000 recent Twinkle user talk edits | link to template |
| Uw-vandalism1 | 226 | [1] |
| Uw-vandalism2 | 124 | [2] |
| Uw-vandalism3 | 115 | [3] |
| Uw-vandalism4 | 57 | [4] |
| Uw-block | 43 | [5] |
| Uw-unsourced1 | 41 | [6] |
| Uw-test1 | 40 | [7] |
| Uw-delete1 | 37 | [8] |
| Uw-spam1 | 28 | [9] |
| Uw-delete2 | 25 | [10] |
| Uw-vandalism4im | 23 | [11] |
| Uw-error1 | 17 | [12] |
| Uw-speedy1 | 17 | [13] |
| Uw-vblock | 17 | [14] |
| Uw-speedy2 | 15 | [15] |
| Uw-unsourced2 | 14 | [16] |
| Uw-unsourced3 | 14 | [17] |
| Uw-editsummary | 13 | [18] |
| Uw-spam2 | 13 | [19] |
| Uw-test2 | 12 | [20] |
The final counts delivered through this method is not completely accurate, because any other, previous messages on the user talk pages these revisions were pulled from which ALSO used one of the listed user warning templates (whether or not it was left using Twinkle), will also be counted. There may not be an easy way around this using the API Query. There is a way to pull up just the text of a single diff by passing non-Query API arguments, but we couldn't make that method work this time. However, this count is probably pretty close. And after all, any user warning template that is in active use is a good candidate for A/B testing, whether or not it's used solely by Twinklers.
Frequency of Twinkle use
We also ran a query to see how many Twinkle user talk edits there had been in recent months. According to this query, there have been 161,934 Twinkle messages posted on user talk pages between the beginning of July and October 19th 2011. Next, we intend to run the same query for Huggle and Igloo edits and to count total user talk page edits during the same time period, in order to find out which of these tools is used most frequently on user talk pages, and what percentage of all user talk page messages are made using these tools.