- We are now tracking when contributors on mobile and desktop switch from one editing interface to another
|mediawiki/extensions/WikiEditor||master||+116 -65||Log editor switches to visualeditorfeatureuse|
|mediawiki/extensions/VisualEditor||master||+38 -1||Log editor switches to visualeditorfeatureuse|
|mediawiki/extensions/MobileFrontend||wmf/1.34.0-wmf.11||+25 -4||Log editor switches to visualeditorfeatureuse|
|mediawiki/extensions/MobileFrontend||master||+25 -4||Log editor switches to visualeditorfeatureuse|
|Open||None||T255327 [Epic] Evaluate which editing interface should be shown by default|
|Open||None||T227338 Test visual editor as the default mobile editor on select wikis|
|Resolved||DLynch||T221191 VE mobile default: start tracking editor switches on mobile + desktop|
@Neil_P._Quinn_WMF That's something which has been subject to general code-atrophy over time. It only fires the abort for a full teardown editor-switch on desktop (so NWE isn't counted), and mobile has gradually wound up getting in-line with that so it's not logging the aborts because it's not navigating away. I figured putting in a consistent "this is feature-usage, and not an abstract thing tied to page lifecycle" would get us a less fragile metric going forward.
I found an extra little data-glitch in there -- when switching from WikiEditor to VisualEditor it wouldn't log it properly, because it wasn't updated to correctly see the in-page load. It would thus wait to log an abort until VE was done and triggered a page-reload, and it wouldn't log that abort correctly as being a switch away from WikiEditor (because it detected that purely based on whether the link to visual editing currently had focus).
So, @Neil_P._Quinn_WMF: all analytics data about switching to VE from WikiEditor from the last 3-ish years is suspect. 😁
Hahaha, wow, multiple good catches! This will definitely help us tracking switching behavior during the A/B test 👏
Out of curiosity, I checked the percentage of aborts in each editor that were switches, and it lines up with what you say (where desktop VE is the only editor logging any at all).
select event.platform as platform, event.editor_interface as editor, concat(round(( sum(cast(event.abort_type like "switch%" as int)) * 100 / count(*) ), 1), "%") as switch_aborts -- From the unsanitized data, so it only covers the last 90 days from event.editattemptstep where event.action = "abort" and -- Partition predicate to make Hive happy year > 0 group by event.platform, event.editor_interface