On a MediaWiki install we have special pages (https://www.wikidata.org/wiki/Special:SpecialPages) like Most linked-to pages and Pages with the most interwikis. The wikibase extension should add an extension for items with the most sitelinks.
Rough query:
SELECT ips_item_id, COUNT(ips_item_id) FROM wb_items_per_site GROUP by ips_item_id ORDER BY COUNT(ips_item_id) DESC LIMIT 100;
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Implement a special page to show items with the most sitelinks | mediawiki/extensions/Wikibase | master | +149 -1 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T96040 Wikibase special pages (tracking) | |||
| Declined | Ricordisamoa | T48217 Implement a special page to show items with the most sitelinks | |||
| Resolved | daniel | T42157 Store hints about entities in page_props table | |||
| Declined | None | T60028 cache special page results | |||
| Resolved | None | T60032 Allow page_props to be queried by value. | |||
| Resolved | • Springle | T66411 Deploy pp_sortkey schema change to all wikis |
Maarten: You mean we should have such a special page on Wikidata? Or on the client (Wikipedia, Wikivoyage, Commons, ...)?
Change 94830 had a related patch set uploaded by Bene:
(bug 46217) Implement a special page to show items with the most sitelinks
The suggested patch needs to be changed to allow more efficient SQL (see bug 40157 and bug 58032).
Change 94830 abandoned by Bene:
(bug 46217) Implement a special page to show items with the most sitelinks
Reason:
this is not likely to get implemented this way
Change 181902 had a related patch set uploaded (by Multichill):
Implement a special page to show items with the most sitelinks
I took a different approach than Bene: I extended the standard QueryPage. Would appreciate input. If this approach works I plan to add some more special pages and maybe rewrite Special:unconnectedpages.
Change 181902 had a related patch set uploaded (by Siebrand):
Implement a special page to show items with the most sitelinks
Change 181902 abandoned by Multichill:
Implement a special page to show items with the most sitelinks
Reason:
Too frustrating, not going to invest time in this any more.
The major problem with both patches is that they use COUNT and JOIN and GROUP BY. This does not scale well for such big tables. What we do now is to store the number of sitelinks per item in a "wb-sitelinks" page property (in the page_props table). pp_sortkey is a numeric field that can be accessed then, and possibly ordered (need to check this). This should already be deployed, as far as I know.
So whoever wants to pick this up, please pick one of the existing patches, reopen it and change the SQL query to query pp_sortkey in page_props instead.
Change 181902 restored by JanZerebecki:
Implement a special page to show items with the most sitelinks
Reason:
Restoring to make it possible for Ricordisamoa to work on it.
With https://gerrit.wikimedia.org/r/232698 I explore a different approach: it would be possible to use something like https://www.wikidata.org/wiki/Special:PagesWithProp/wb-sitelinks?sortbyvalue=1.
The same would work with wb-claims, etc.
The page property wb-sitelinks got introduced. That might open some options. It's stored as a string so no simple output
MariaDB [wikidatawiki_p]> SELECT * FROM page_props WHERE pp_propname='wb-sitelinks' ORDER BY pp_value DESC LIMIT 20; +---------+--------------+----------+------------+ | pp_page | pp_propname | pp_value | pp_sortkey | +---------+--------------+----------+------------+ | 11009 | wb-sitelinks | 99 | 99 | | 28674 | wb-sitelinks | 99 | 99 | | 43781 | wb-sitelinks | 99 | 99 | | 6406 | wb-sitelinks | 99 | 99 | | 6151 | wb-sitelinks | 99 | 99 | | 28679 | wb-sitelinks | 99 | 99 | | 6271239 | wb-sitelinks | 99 | 99 | | 28680 | wb-sitelinks | 99 | 99 | | 8267529 | wb-sitelinks | 99 | 99 | | 780 | wb-sitelinks | 99 | 99 | | 6307853 | wb-sitelinks | 99 | 99 | | 5741069 | wb-sitelinks | 99 | 99 | | 8838925 | wb-sitelinks | 99 | 99 | | 51214 | wb-sitelinks | 99 | 99 | | 10768 | wb-sitelinks | 99 | 99 | | 14868 | wb-sitelinks | 99 | 99 | | 28698 | wb-sitelinks | 99 | 99 | | 6674203 | wb-sitelinks | 99 | 99 | | 3614 | wb-sitelinks | 99 | 99 | | 7710 | wb-sitelinks | 99 | 99 | +---------+--------------+----------+------------+ 20 rows in set (2 min 50.05 sec)
This change got very old and in the meantime we have this information in SPARQL so anyone can retrieve this data. I don't think this is needed any more. I propose to abandon the change and close the task.
Change 181902 abandoned by Thiemo Mättig (WMDE):
Implement a special page to show items with the most sitelinks
@thiemowmde: Ricordisamoa was working on this change so I didn't close it yet because I wanted him to have the chance to comment. @Ricordisamoa sorry for the quick close.