Package org.apache.jena.sparql.algebra.optimize

Class TransformOrderByDistinctApplication

  • All Implemented Interfaces:
    Transform
    public class TransformOrderByDistinctApplication
extends TransformCopy

    Improved optimization for ORDER BY plus DISTINCT or REDUCED combinations, see JENA-441 for original proposal and discussion.

    This optimization is enabled by default as with most ARQ optimizations and may be disabled by setting the symbol ARQ.optOrderByDistinctApplication to false.

    Optimization Applicability

    This is a limited optimization that applies in the case where you have a query that meets the following conditions:

    • Uses both ORDER BY and DISTINCT or REDUCED on the same level of the query
    • There is a fixed list of variables to project i.e. not SELECT *
    • ORDER BY conditions only use variables present in the project list

    Essentially this takes algebras of the following form: 

     (distinct 
   (project (?var) 
     (order (?var) 
       ... )))

    And produces algebra of the following form: 

     (order (?var)
   (distinct 
     (project (?var) 
       ... )))

    In the general case this in unsafe because it would change the semantics of the query since ORDER BY can access variables that are not projected. However if the conditions outlined are met then this optimization is safe, the algebras will be semantically equivalent and the resulting form likely significantly more performant, of course YMMV depending on how much data you are querying.