Page MenuHomePhabricator

feature request: ability to exclude a wiki from search results
Open, Needs TriagePublicFeature

Description

We have some copy and paste bots that appear and copy articles from English Wikipedia to the small wikipedias, pasting being in English. Bots being unrequested, and resulting in the deletion of said pages.

Trying to identify which of the small wikipedias they are appearing next is a little difficult as you can get swamped by the enwiki results. If we could exclude a wiki, eg -enwiki: -commonswiki that would be useful

Thanks for the consideration, for those who do off-piste wiki searches. :-)

Event Timeline

Billinghurst renamed this task from featuer request: ability to exclude a wiki from search results to feature request: ability to exclude a wiki from search results.May 10 2020, 5:48 AM

Looking at the source, I think this might be as easy as changing the way that the q data is sent to ElasticSearch from using a match against source_text.plain to a simple_query_string. That would allow using expressions like AND NOT wiki:enwiki in the search. I don't know however what core functionality of the tool this might break. It almost certainly wouldn't pair well with the 'Regular expression' option.

If the desire was to limit the results on only one wiki, the best fix would probably be changing from the *,*:* cross cluster search index URL component (which means roughly "all local indices, all clusters:all indices") to something that looked up the right cluster to target based on the selected wiki and then only routed to that single <cluster>:<index> target.

Thinking a little broadly rather than my initial issues for looking for problematic editing outside of one wiki, to what and how one might wish to quickly limit/focus searches.

  • language based — show me those wikis where the native language is Spanish
  • sister/domain based — show me these results across all wiktionaries or all wikimedia domains
1234qwer1234qwer4 changed the subtype of this task from "Task" to "Feature Request".Jun 15 2021, 2:04 PM
1234qwer1234qwer4 subscribed.