3-4% of save success events in the EditAttemptStep data stream contain the same revision ID as previous events in the session. Since saving a new revision must generate a new revision ID, this indicates that something is wrong with those events.
We do not frequently use the revision_id field, and the proportion of affected events is small, so this is not urgent.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T118063 Clean up the EditAttemptStep schema and its implementations | |||
| Resolved | MNeisler | T231024 Some save success events do not contain a new revision ID |
Interesting idea! But the desktop visual editor and desktop wikitext editor are equally affected, and unless I'm mistaken, you can't make a null edit in VE.
We don't enable the "Publish" button until you make changes, but nothing stops you from publishing after typing and then deleting a space somewhere. That will result in a null edit.
I just checked (with ?trackdebug=1) and the mwedit.saveSuccess event we generate in that case has a revision_id field that corresponds to the latest revision of the page.
Does this explain the data you were seeing?
Hey @MNeisler can you help us move this task along? We would like to know the consequences of this. CC: @Neil_P._Quinn_WMF
I discussed this with @Neil_P._Quinn_WMF. @matmarex's explanation makes sense and these are likely NULL edits.
I've updated the description of the revision_id field in the EditAttemptStepSchema to clarify that NULL edits will have the same revision ID as previous events in the session. I think we can go ahead and mark this task as resolved. Thanks all!