The 2010 wikitext editor has not been emitting init, saveAttempt, saveSuccess, or saveFailure events. It also has not been emitting saveIntent events, but this is intentional since the editor's flow has no step that corresponds to the click to bring up the save dialog in the visual editor.
Description
Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | nshahquinn-wmf | T202132 EPIC: Generate one-off metric snapshots for mobile editing documentation | |||
Resolved | nshahquinn-wmf | T202147 Snapshot: which parts of the edit process cause people to abort edits? | |||
Resolved | nshahquinn-wmf | T202437 Identify and fix data quality problems in the Edit event log | |||
Resolved | Jdforrester-WMF | T203621 2010 wikitext editor logs init and save events to old schema |
Event Timeline
Same reasoning as it not logging an init event, maybe? Insofar as there's not an in-page equivalent to the saveComplete event that VE emits, because there's a page-navigation in the way.
@Jdforrester-WMF You chimed in on the parent ticket about the reasoning for init -- can you remember if that was our reasoning for this as well?
@Neil_P._Quinn_WMF Is there some server-side logging via the edit session that you can use for an equivalent measure.
Wait, but then how did we ever calculate the edit save rate for the 2010 wikitext editor?
@Neil_P._Quinn_WMF Is there some server-side logging via the edit session that you can use for an equivalent measure.
Well, I was told (I think by @Milimetric, years ago) that init and saveSuccess were logged on the server-side. But I don't entirely understand the question: are you asking whether I can use data from the revision history to infer saveSuccess? That seems pretty difficult and much less desirable than logging it to this schema on the server.
@Jdforrester-WMF confirmed to me that he was referring to the saveIntent events, which are intentionally not emitted. The other save events should indeed be emitted (some or all on the server side), so that is a real bug.
Hmmmm, the queries / data / graphs / my recollection all seem to point to attempt, success, and failure being saved, for wikitext: https://edit-analysis.wmflabs.org/compare/
But looking at that saves_rates query, it never differentiated by editor, like here: https://phabricator.wikimedia.org/diffusion/ANLD/browse/master/edit/sessions.sql$13
But then, yeah, if sessions.sql *did* differentiate by editor, and it seems to have thought saveAttempt was one of the steps in the wikitext flow, where is that data? Because if I look for saveAttempt now, it's not there in Edit_10676603 in mysql. In fact no event_editor other than 'visualeditor' shows up there. So that doesn't make sense. The hive event.edit table is empty, because of the issue it had with the automatic refine process, so it's not there... And I don't see how only part of the data would've been sanitized...
So it looks like Edit_11319708 and onward have this data, @Neil_P._Quinn_WMF when you have a sec let me know what you're looking for exactly. I think it's weird that the older tables don't have wikitext but I think I remember it wasn't instrumented back then?
Are you sure? I queried the most recent table (Edit_17541122) for August 13 to August 27, and found no save events of any kind for the 2010 wikitext editor.
My notebook is here—if you go down to the "overall event counts" section, there's a pivot table showing the count of events by editor. Am I missing something?
Change 460618 had a related patch set uploaded (by Jforrester; owner: Jforrester):
[mediawiki/extensions/WikiEditor@master] Follow-up 874f9bfb8d: Bump server-side Edit schema version here
Yup, it was that simple—both init and save events are logged server-side, and were just logging to an old schema version all this time.
Change 460618 merged by jenkins-bot:
[mediawiki/extensions/WikiEditor@master] Follow-up 874f9bfb8d: Bump server-side Edit schema version here
This is now fixed! Starting on the day of the deploy, these events started flowing into the new table and stopped flowing into the old one.
select left(timestamp, 8) as day, count(*) from log.Edit_17541122 where event_editor = "wikitext" and event_action in ("init", "saveAttempt", "saveSuccess", "saveFailure") and timestamp >= "20180911" group by left(timestamp, 8) day count(*) 0 20180912 2 1 20180913 2 2 20180914 10 3 20180915 1 4 20180917 1 5 20180918 20 6 20180919 4273 7 20180920 73329 8 20180921 144332 9 20180922 133908 10 20180923 144711 11 20180924 104613
select left(timestamp, 8) as day, count(*) from log.Edit_13457736 where event_editor = "wikitext" and event_action in ("init", "saveAttempt", "saveSuccess", "saveFailure") and timestamp >= "20180911" group by left(timestamp, 8) day count(*) 0 20180911 154631 1 20180912 149683 2 20180913 153639 3 20180914 136980 4 20180915 130795 5 20180916 140776 6 20180917 149348 7 20180918 137878 8 20180919 136892 9 20180920 69104 10 20180923 1