Check the usage data in Hadoop.
|Resolved||Lucas_Werkmeister_WMDE||T255880 Remove uses of PHP classes from Wikibase Repo in Client and Lib|
|Resolved||Lucas_Werkmeister_WMDE||T255882 Don’t use WikibaseRepo when initializing Client pageterms API|
|Resolved||Ladsgroup||T256255 Investigate how pageterms API is used on Wikidata itself|
SELECT `database`, COUNT(*) as hitcount FROM event.mediawiki_api_request WHERE year = 2020 AND month = 6 AND day = 14 and params['action'] = 'query' and params['prop'] = 'pageterms' group by `database`; database hitcount ukwiki 1 ptwiki 482 rowiki 13 trwikinews 2 wikidatawiki 338 nlwiki 2 fawiki 45 svwiki 9 cswiki 2 zhwiki 10 plwiki 34 enwiki 927 ruwiki 120 eswiki 243 jawiki 19 frwiki 188 dewiki 62 arwiki 33 itwiki 67
Paramaters sent to cases of wikidatawiki requests can be found in P11653 (Includes private data, protected for view but you should have access)
I thought about this some more. We have basically 4 cases I can see:
- WP article is connected to a WD Item. We call the API on the WP article. We get back the data for the WD Item. ✅
- WD non-Item page is connected to a WD Item. We call the API on the WD non-Item page. We get back the data for the WD Item. ✅
- WD Item is connected to another WD Item. We call the API on the first. What should we get back? 🚫
- WD Item exists and we call the API directly on it, not through its sitelink connection. What should we get back? 🚫
3 is an outlier we declare to be handled like 1 and 2. 1 and 2 are what we intended it for. I fear people are using it for 4. Can we confirm this by looking at some sample calls for the API? If that assumption is correct can we deprecate that usage and point people at an alternative?
Anything I'm missing?