Plugins » Katello

In https://github.com/Katello/katello/blob/master/app/models/katello/repository.rb#L291, errata_with_packages is a set with many duplicates - this can be visible by enabling sql debugs, publishing a CV and grepping for "katello_errata.id NOT IN" in production.log.

Does it make sense to make the set uniq before passing it as "NOT IN" option to postgres? Wont it improve performance? I mean trivial patch like:

--- repository.rb    2017-12-26 14:49:55.133400485 +0100
+++ repository.rb.new    2017-12-26 14:51:39.772839567 +0100
@@ -293,7 +293,7 @@ module Katello
         where("#{repository_errata}.repository_id" => self.id)

       if errata_with_packages.any?
-        self.errata.where("#{Katello::Erratum.table_name}.id NOT IN (?)", errata_with_packages.pluck("#{errata}.id"))
+        self.errata.where("#{Katello::Erratum.table_name}.id NOT IN (?)", errata_with_packages.pluck("#{errata}.id").uniq)
       else
         self.errata
       end

I tried this on my foreman/katello but got bit contradicting results so far - sometimes the uniq made by ruby takes overall command faster, sometimes the original behaviour is faster, times seem to vary a lot.