Page MenuHomePhabricator
Create Task
Maniphest T97543

rev_parent_id not used correctly with diff=prev or rvdiffto=prev for some revisions (Conversion script?)
Closed, DuplicatePublic
Actions

Description

I have a script that is requesting diffs from the API. I've been running into some revisions where the diff that is returned is wrong and I can't figure out why. Let's examine an example revision:

https://en.wikipedia.org/wiki/?oldid=3110 (note that there is no link for "prev" revision)

If I find this revision in the edit history of the page and click on the "prev" link to the left of it, the following link is loaded with the appropriate diff:

https://en.wikipedia.org/w/index.php?title=Hebrew_grammar&diff=3110&oldid=257167

But if I manually specify diff=prev via the UI or API, I don't get any diff at all.

https://en.wikipedia.org/w/index.php?title=Hebrew_grammar&diff=prev&oldid=3110
http://en.wikipedia.org/w/api.php?action=query&prop=revisions&revids=3110&rvdiffto=prev

I originally suspected that rev_parent_id was missing from the revision table, but it seems like it is populated as expected. See http://quarry.wmflabs.org/query/3379 which reports a rev_parent_id of 257167. But note that the previous revision (257167) was performed by "Conversion script"; that might be relevant.

Event Timeline

Halfak created this task.Apr 29 2015, 4:25 PM
Halfak raised the priority of this task from to Needs Triage.
Halfak updated the task description. (Show Details)
Halfak added projects: MediaWiki-Action-API, MediaWiki-Page-diffs.
Halfak subscribed.
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptApr 29 2015, 4:25 PM
Anomie subscribed.Apr 29 2015, 6:55 PM

rev_parent_id isn't used at all for diffs. Actually, I'm not sure what it is used for, or what its exact semantics are when auto-merged edit conflicts are involved.

The source of this issue is that the history page uses the revids from the list of revisions it already has to generate diff links, and that list is ordered by timestamp, while the actual "prev" pseudo-revision used by the diffing code orders by revision id instead. Most of the time this gives the same results, but for revisions that were imported later it doesn't.

Anomie closed this task as a duplicate of T4930: Edit history is based on timestamp while diff prev/next links are based on revision id.Apr 29 2015, 6:55 PM
Aklapper added a project: MediaWiki-Page-history.Sep 20 2019, 7:47 PM
Aklapper removed a subscriber: Anomie.Oct 16 2020, 5:39 PM
Content licensed under Creative Commons Attribution-ShareAlike (CC BY-SA) 4.0 unless otherwise noted; code licensed under GNU General Public License (GPL) 2.0 or later and other open source licenses. By using this site, you agree to the Terms of Use, Privacy Policy, and Code of Conduct. · Wikimedia Foundation · Privacy Policy · Code of Conduct · Terms of Use · Disclaimer · CC-BY-SA · GPL