Page MenuHomePhabricator

purge does not update langlinks table
Closed, ResolvedPublic

Description

Steps to reproduce

  1. SELECT page_namespace, page_title FROM page LEFT JOIN langlinks ON ll_from=page_id WHERE page_namespace=14 AND ll_from IS NULL AND page_title LIKE "Amerikaanse%" LIMIT 3;

This shows there is no entry for "Categorie:Amerikaanse_Maagdeneilanden_op_de_Olympische_Jeugdspelen" in the langlinks table

However, https://nl.wikipedia.org/wiki/Categorie:Amerikaanse_Maagdeneilanden_op_de_Olympische_Jeugdspelen shows there *are* langlinks, provided by WikiData.

Browse to https://nl.wikipedia.org/wiki/Categorie:Amerikaanse_Maagdeneilanden_op_de_Olympische_Jeugdspelen?action=purge

Expected behavior

  • Langlinks table being updated automatically, and lacking that,
  • ?action=purge updating the language table

Actual behavior

  • ?action=purge does *not* update the langlinks entries
  • A null edit does.

Version: master
Severity: critical
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=59962

Details

Reference
bz59961

Event Timeline

bzimport raised the priority of this task from to High.Nov 22 2014, 2:40 AM
bzimport set Reference to bz59961.
bzimport added a subscriber: Unknown Object (MLST).

(In reply to comment #0)

Actual behavior

  • ?action=purge does *not* update the langlinks entries
  • A null edit does.

What if you try using ?action=purge&forcelinkupdate=1 via the API ([[mw:API:Purge]])

It's actually conceptually unclear whether langlinks should contain the *effective* language links, or only the ones actually defined on the wiki page. According to your report, the behavior is actually inconsistent. That would definitely be a bug.

langlinks table should at least not contain langlinks which are neither local defined nor at wikidata repository.

select * from dewiki_p.langlinks where ll_from=6463057;
+---------+---------+-------------+

ll_fromll_langll_title

+---------+---------+-------------+

6463057enMark Atkins
6463057itMark Atkins

+---------+---------+-------------+

http://de.wikipedia.org/wiki/Mark_Atkins (with pageid 6463057) does not contain any local langlinks and on wikidata repository the it sitelink was removed three month ago https://www.wikidata.org/w/index.php?title=Q1899938&diff=128208462&oldid=127387275. So the it-langlink row must be removed.

This is original an update queue problem. But there must be a way to fix this.

Data Analysis Tools have big problems with this incorrect database table data.

Lydia_Pintscher removed a subscriber: Unknown Object (MLST).
Lydia_Pintscher removed a subscriber: Unknown Object (MLST).Dec 1 2014, 2:31 PM
Lydia_Pintscher added subscribers: hoo, aude, JanZerebecki.

Ok so who can decide if this table should contain only locally defined links or also the ones from Wikidata?

hoo claimed this task.

This has been fixed in the mean time, see T89903. Purging without forcelinkupdate is not expected to purge any secondary data, so that's fine (or at least not a Wikidata issue).