BoostStrategy for a Hashset field

veveloko · **Joined:** Wed Mar 27, 2013 8:56 am **Posts:** 1

I have an @Entity User with a @Field that is a HashSet<User>.

I want to know the Document with the max number of users associated to boost with the formula in a CustomBoostStrategy:

Code:

boost = ( User.hashset / MaxNumberFromAnyUser.hashset ) x 0.25;

I can´t Autowired any DAO.

Any help?

Thx :)

antonio.fernandez · **Joined:** Wed Mar 13, 2013 8:41 am **Posts:** 2

Hi all,

This is my first post in this forum, and I would first like to thank the whole team of Hibernate-Search for the great work done with this framework.

By reading this post I think @ veveloko has set a feature that I too may be interested. I think he wants to instantiate the class CustomBoosStrategy from Spring IoC container, in order to inject dependencies as a DAO class using the @Autowired, and finally this class is used by Hibernate-Search for calculating the Boost factor.

After analyzing source code I believe this integration with Spring beans is not possible, because the instantiation of BoosStrategy implementations classs is done by following method of AnnotationProcessinHelper class:

Code:

public static BoostStrategy getDynamicBoost(XProperty member) {
      DynamicBoost boostAnnotation = member.getAnnotation( DynamicBoost.class );
      if ( boostAnnotation == null ) {
         return DefaultBoostStrategy.INSTANCE;
      }

      Class<? extends BoostStrategy> boostStrategyClass = boostAnnotation.impl();
      BoostStrategy strategy;
      try {
         strategy = boostStrategyClass.newInstance();
      }
      catch ( Exception e ) {
         throw new SearchException(
               "Unable to instantiate boost strategy implementation: " + boostStrategyClass.getName()
         );
      }
      return strategy;
   }

Overall I would like to know if I can use Spring beans to create my implementations of Similarity, BoosStrategy, etc. ..

Thanks in advance, have a nice day
Antonio

sanne.grinovero · **Posted:** Wed Apr 03, 2013 7:08 am

Antonio: you seem to have a more general question so let's discuss that in your other thread.

veveloko:

conceptually the problem is that while "MaxNumberFromAnyUser.hashset" might be a fixed number when you index this specific entity, the value might change after you save other entities. What would you want to happen in that case?
With such a solution you might need to rebuild the whole index at every database write, which is going to be extremely slow.

My advice is to use boosting at query time: you can affect the score on each document when running a query rather than write the boost score at index time.

See also org.apache.lucene.search.function.CustomScoreQuery