Robert, is that feasible for the Lucene index?

To do it in the MySQL backend we'd need to add those fields to searchindex (or else build a temporary table w/ results -- anything to avoid joining searchindex and page which is the Performance Kiss Of Death) and I'm still not sure how efficient that would run.

rd232 wrote:

Any update on feasibility of sorting search results?

nilesh wrote:

Can Lucene 3.4.0 (at least) be used with MediaWiki? Are there any problems/limitations?

It has support for index-time joins that denormalize relational database tables so that we can simply search over those tables in real time without any joining overhead. Please check the docs for the same for v 3.6.0 here - http://lucene.apache.org/core/3_6_0/api/contrib-join/org/apache/lucene/search/join/package-summary.html

What do you think about using it?

(In reply to comment #3)

Can Lucene 3.4.0 (at least) be used with MediaWiki? Are there any
problems/limitations?

https://wikitech.wikimedia.org/wiki/Search has info about the search infra and what's used with what. Let's please keep this bug report specific and focused.

Added project CirrusSearch since it's the new search engine used by wikimedia wikis. Also added number of words as criteria since it is also shown on search results.

Just noticed T11519, I'll fill a separate bug for date of latest modification, and another for multimedia search options.

We've added some of this sorting capability since this was filed, but page size is not among those updates. Since that's captured in another ticket, we'll close this one.

See the sort parameter to the search api, added in the last few months: https://en.wikipedia.org/wiki/Special:ApiSandbox#action=query&format=json&list=search&srsearch=qqq&srinterwiki=1&srsort=relevance

Available sorts include last edited date, incoming link counts and relevance. Soon also sorting by creation date. We could plausibly add sorting by word count but we would need to expand on the use case.