Some of our metrics would benefit from more normalization. For example, Visual Editor template dialog opens make the most sense as a rate per Visual Editor session. Discuss any open questions and use aggregation scripts to produce any missing denominators.
If aggregation scripts will be adjusted, there is a help page and the queries are in this repo.
Wanted normalizations
| Context | Goal | Wanted | Implementation |
| CodeMirror (board, config.yaml) | Percentage of code mirror users per editor type | All users of a editor type (byEditor) | Read editor types and codemirror settings from user_properties table → gerrit:657362 |
| Visual Editor TemplateDialog (board, config.yaml) | Percentage of “open” actions per VE pageview (session) | Total number VE sessions by edit count (Note that these are both sampled at 8.25%) | EditAttemptStep with editor_interface (visualeditor, wikitext-2017) & action (ready) & user_editcount → gerrit:659230 |
Combinations of relevant preference settings (CodeMirror)
| Wikitext 2010 usebetatoolbar | Wikitext 2017 visualeditor-newwikitext | Code mirror usecodemirror | Outcome |
| 0 | (0) | 0 (or missing) | disabled (2010 editor) |
| 0 | (0) | 1 | disabled (2010 editor) |
| 0 | 1 | 0 (or missing) | disabled (2017 editor) |
| 0 | 1 | 1 | enabled (2017 editor) |
| (1) | (0) | 0 (or missing) | disabled (2010 editor) |
| (1) | (0) | 1 | enabled (2010 editor) |
| (1) | 1 (precedence) | 0 (or missing) | disabled (2017 editor) |
| (1) | 1 (precedence) | 1 | enabled (2017 editor) |
The gadget wikiEd takes precedence over the 2010 editor but not over 2017 editor.