SeamFramework.orgCommunity Documentation

Chapter 29. Hibernate Search

29.1. Introduction
29.2. Configuration
29.3. Usage

Full text search engines like Apache Lucene™ are a very powerful technology that bring full text and efficient queries to applications. Hibernate Search, which uses Apache Lucene under the covers, indexes your domain model with the addition of a few annotations, takes care of the database / index synchronization and returns regular managed objects that are matched by full text queries. Keep in mind, thought, that there are mismatches that arise when dealing with an object domain model over a text index (keeping the index up to date, mismatch between the index structure and the domain model, and querying mismatch). But the benefits of speed and efficiency far outweigh these limitations.

Hibernate Search has been designed to integrate nicely and as naturally as possible with JPA and Hibernate. As a natural extension, JBoss Seam provides an Hibernate Search integration.

Please refer to the Hibernate Search documentation for information specific to the Hibernate Search project.

Hibernate Search is configured either in the META-INF/persistence.xml or hibernate.cfg.xml file.

Hibernate Search configuration has sensible defaults for most configuration parameters. Here is a minimal persistence unit configuration to get started.


<persistence-unit name="sample">
   <jta-data-source>java:/DefaultDS</jta-data-source>
   <properties>
      [...]
      <!-- use a file system based index -->
      <property name="hibernate.search.default.directory_provider" 
         value="org.hibernate.search.store.FSDirectoryProvider"/>
      <!-- directory where the indexes will be stored -->
      <property name="hibernate.search.default.indexBase" 
         value="/Users/prod/apps/dvdstore/dvdindexes"/>
   </properties>
</persistence-unit>

If you plan to target Hibernate Annotations or EntityManager 3.2.x (embedded into JBoss AS 4.2.x and later), you also need to configure the appropriate event listeners.


<persistence-unit name="sample">
   <jta-data-source>java:/DefaultDS</jta-data-source>
   <properties>
      [...]
      <!-- use a file system based index -->
      <property name="hibernate.search.default.directory_provider" 
                value="org.hibernate.search.store.FSDirectoryProvider"/>
      <!-- directory where the indexes will be stored -->
      <property name="hibernate.search.default.indexBase" 
                value="/Users/prod/apps/dvdstore/dvdindexes"/>

      <property name="hibernate.ejb.event.post-insert" 
                value="org.hibernate.search.event.FullTextIndexEventListener"/>
      <property name="hibernate.ejb.event.post-update" 
                value="org.hibernate.search.event.FullTextIndexEventListener"/>
      <property name="hibernate.ejb.event.post-delete" 
                value="org.hibernate.search.event.FullTextIndexEventListener"/>
                
   </properties>
</persistence-unit>

In addition to the configuration file, the following jars have to be deployed:

Hibernate Search uses annotations to map entities to a Lucene index, check the reference documentation for more informations.

Hibernate Search is fully integrated with the API and semantic of JPA / Hibernate. Switching from a HQL or Criteria based query requires just a few lines of code. The main API the application interacts with is the FullTextSession API (subclass of Hibernate's Session).

When Hibernate Search is present, JBoss Seam injects a FullTextSession.

@Stateful

@Name("search")
public class FullTextSearchAction implements FullTextSearch, Serializable {
   @In FullTextSession session;
   public void search(String searchString) {
      org.apache.lucene.search.Query luceneQuery = getLuceneQuery();
      org.hibernate.Query query session.createFullTextQuery(luceneQuery, Product.class);
      searchResults = query
            .setMaxResults(pageSize + 1)
            .setFirstResult(pageSize * currentPage)
            .list();
   }
   [...]
}

Note

FullTextSession extends org.hibernate.Session so that it can be used as a regular Hibernate Session

If the Java Persistence API is used, a smoother integration is proposed.

@Stateful

@Name("search")
public class FullTextSearchAction implements FullTextSearch, Serializable {
   
   @In FullTextEntityManager em;
   public void search(String searchString) {
      org.apache.lucene.search.Query luceneQuery = getLuceneQuery();
      javax.persistence.Query query = em.createFullTextQuery(luceneQuery, Product.class);
      searchResults = query
            .setMaxResults(pageSize + 1)
            .setFirstResult(pageSize * currentPage)
            .getResultList();
   }
   [...]
}

When Hibernate Search is present, a FulltextEntityManager is injected. FullTextEntityManager extends EntityManager with search specific methods, the same way FullTextSession extends Session.

When an EJB 3.0 Session or Message Driven Bean injection is used (i.e. via the @PersistenceContext annotation), it is not possible to replace the EntityManager interface by the FullTextEntityManager interface in the declaration statement. However, the implementation injected will be a FullTextEntityManager implementation: downcasting is then possible.

@Stateful

@Name("search")
public class FullTextSearchAction implements FullTextSearch, Serializable {
  
   @PersistenceContext EntityManager em;
   public void search(String searchString) {
      org.apache.lucene.search.Query luceneQuery = getLuceneQuery();
      FullTextEntityManager ftEm = (FullTextEntityManager) em;
      javax.persistence.Query query = ftEm.createFullTextQuery(luceneQuery, Product.class);
      searchResults = query
            .setMaxResults(pageSize + 1)
            .setFirstResult(pageSize * currentPage)
            .getResultList();
   }
   [...]
}

Caution

For people accustomed to Hibernate Search out of Seam, note that using Search.getFullTextSession is not necessary.

Check the DVDStore or the blog examples of the JBoss Seam distribution for a concrete use of Hibernate Search.