What's requested: Re-do https://commons.wikimedia.org/wiki/File:2018-10_Wikimedia_editing_interface_retention.png#Summary to only count edits to the mainspace.
Why it's requested:
- Uploading files is none of these edit types, and it's not clear how they were counted. Commons generates about 1 out of 6 "edits" each month, so this could be a considerable factor. There are relatively few plain-old-editing edits at Commons: it's either file uploading or script-based edits (e.g., HotCat, which gets counted as a "wikitext edit" even if you never see wikitext, because T188433: Tag all edits by editor used (WikiEditor, API, etc), not just VE hasn't been implemented yet). None of those should be counted as "wikitext edits". It therefore might be easiest to limit this test to Wikipedias only.
- The visual editor is enabled only in a few namespaces, and reaching anything except the mainspace is difficult in the mobile platforms. The retention of "people who started in the mainspace on mobile" vs "people who started in any namespace in wikitext" doesn't feel like an apples-to-apples comparison. Therefore, I suggest doing this for people editing the mainspace only.
- Then there's the problem of the visual editor being configured differently at different wikis, e.g., "hidden" for some users at the English Wikipedia. So perhaps separate numbers for English Wikipedia vs other Wikipedias would be a way to determine if config differences matter to retention. The "Phase 7" Wikipedias (at https://www.mediawiki.org/wiki/VisualEditor/Rollouts ) is another group to consider for separation.
When it's requested: As soon as reasonably convenient.