Hibernate Search is a library that integrates Hibernate ORM with Apache Lucene or Elasticsearch by automatically indexing entities, enabling advanced search functionality: full-text, geospatial, aggregations and more. For more information, see Hibernate Search on hibernate.org.

We just published Hibernate Search version 5.8.0.Beta3, with bugfixes and improvements over 5.8.0.Beta2, but also new features such as analyzer providers, normalizers, AWS compatibility and SPIs for integration of dependency injection frameworks!

Hibernate Search 5.8.x, just as 5.7.x, is only compatible with Hibernate ORM 5.2.3 and later.

If you need to use Hibernate ORM 5.0.x or 5.1.x use the older Hibernate Search 5.6.x.

About 5.8

Hibernate Search 5.8 is mainly about:

  • making the Elasticsearch integration compatible with Elasticsearch 5.x (done);

  • improving performance of the Elasticsearch integration (in progress);

  • introducing a new DSL for defining analyzers (done);

  • ensuring that Hibernate Search will work with Java 9 (done, though Java 9 may still change);

  • improving and documenting WildFly Swarm integration (in discussion);

  • removing the need for class definition on master nodes in JMS/JGroups integration (in discussion);

  • and of course, fixing reported bugs.

You can have a look at the roadmap for more details.

What’s new since Beta2?

Programmatic analyzer definitions

You can now define your analyzers programmatically (without annotations), globally (without putting the definition on a particular entity), and in a native way (without using Lucene classes to configure an Elasticsearch analyzer) using analyzer definition providers.

For example, for Lucene your LuceneAnalysisDefinitionProvider might look like this:

public static class CustomAnalyzerProvider implements LuceneAnalysisDefinitionProvider {
    public void register(LuceneAnalyzerDefinitionRegistryBuilder builder) {
                .analyzer( "myAnalyzer" )
                        .tokenizer( StandardTokenizerFactory.class )
                        .charFilter( MappingCharFilterFactory.class )
                                .param( "mapping", "org/hibernate/search/test/analyzer/mapping-chars.properties" )
                        .tokenFilter( ASCIIFoldingFilterFactory.class )
                        .tokenFilter( LowerCaseFilterFactory.class )
                        .tokenFilter( StopFilterFactory.class )
                                .param( "mapping", "org/hibernate/search/test/analyzer/stoplist.properties" )
                                .param( "ignoreCase", "true" );

While for Elasticsearch you would have:

public static class CustomAnalyzerProvider implements ElasticsearchAnalysisDefinitionProvider {
    public void register(ElasticsearchAnalysisDefinitionRegistryBuilder builder) {
        builder.analyzer( "tweet_analyzer" )
                .withTokenizer( "whitespace" )
                .withCharFilters( "custom_html_strip" )
                .withCharFilters( "p_br_as_space" );

        builder.charFilter( "custom_html_strip" )
                .type( "html_strip" )
                .param( "escaped_tags", "br", "p" );

        builder.charFilter( "p_br_as_space" )
                .type( "pattern_replace" )
                .param( "pattern", "<p/?>|<br/?>" )
                .param( "replacement", " " )
                .param( "tags", "CASE_INSENSITIVE" );

As you can see, this allows you to avoid needing to refer to Lucene classes to configure Elasticsearch analyzers.

More details can be found here for Lucene and here for Elasticsearch.

Normalizers for safer sorts

In HSEARCH-2726 and HSEARCH-2659 we introduced normalizers: analyzers that do not perform any kind of tokenization.

We shamelessly borrowed this concept from Elasticsearch, but implemented it in both embedded Lucene mode and Elasticsearch mode. Normalizers are useful for sortable fields: when a field is sortable it should never be tokenized, as this would make the sort order unpredictable; the sort could apply to the first token if you’re lucky, but it could be applied on any other token.

From version 5.8.0.Beta3 onwards, Hibernate Search will log warnings whenever you’re using an analyzer on a sortable field. To resolve this warning change your Analyzer definition to be a Normalizer.

In Lucene, normalizers are just here to help, they work exactly as analyzers. The two differences are that you can’t affect a tokenizer to a normalizer when defining it, and that normalizers have a runtime safety net: should you manage to create multiple tokens, Hibernate Search will concatenate them back to a single token and log a warning.

In Elasticsearch version 5.2 and above, a normalizer will be translated to a native Elasticsearch normalizer, and a text field with a normalizer will take the keyword datatype.

In Elasticsearch version 5.1 and below, native normalizers are not available, thus normalizers are simply translated to analyzers and a text field with a normalizer will take the text (5.x) or string (2.x) datatype.

You can find out more about normalizers in the reference documentation:

AWS compatibility

AWS requires specific, dynamically computed headers in HTTP requests to handle authentication, which until now has made it difficult to use Hibernate Search with an AWS-hosted Elasticsearch.

We introduced a new SPI allowing low-level configuration of the HTTP client, which allows you to plug in the code required to perform the required AWS authentication; this same SPI may be used to integrate with other cloud providers.

We currently have all of our test suite running successfully against an Elasticsearch cluster managed by AWS, with security turned on.

At this stage the SPI is available but we didn’t release the signing component yet; this will be availble in the next milestone: see introduce an AWS module if you want to help!

Dependency injection in FieldBridges

As part of HSEARCH-1316, we’re experimenting with integration with various dependency injection frameworks.

The integration is about allowing you to use annotations such as @Inject, @PostConstruct and so on in your FieldBridges, which may for example allow you to fetch additional data from your application when indexing a given bean.

Integration is currently known to work with Spring DI and CDI. We don’t provide packages for user consumption, but if you are an integrator, or simply if you feel like it, you can have a look at our integration tests:

And more!

A summary of other notable changes:

  • HSEARCH-2606: the discovery_scheme configuration property is now correctly taken into account. Thanks to Matthieu Vincent for reporting and fixing this issue!

  • HSEARCH-2477: shard filtering now works on Elasticsearch.

  • HSEARCH-2603: we now use the Painless scripting language when doing spatial searches on Elasticsearch 5+. Incidentally, this means that it is no longer necessary to perform any server-side configuration on Elasticsearch 5+ to perform any spatial query.

  • HSEARCH-2734: due to a lot of confusion and incorrect (harmful) use, we have deprecated the "ram" name for the RAMDirectory directory provider. If you need it, please ensure you are not using it in a production environment, read about its limitations in the reference documentation, and use its new name: "local-heap".

  • HSEARCH-2735: index-time boosting features (@Boost, @DynamicBoost) have been deprecated with no replacement, and will need to be removed in a future version because Lucene 7 won’t allow index-time boosting anymore. See the reference documentation for alternatives: the suggestion is to switch to using query-time boosting instead.

  • HSEARCH-2665: IndexingInterceptor is no longer considered experimental.

  • HSEARCH-2666: IndexControlMBean is no longer considered experimental.

For a full list of changes since 5.8.0.Beta2, please see the release notes.

How to get this release

All versions are available on Hibernate Search’s web site.

Ideally use a tool to fetch it from Maven central; these are the coordinates:


To use the experimental Elasticsearch integration you’ll also need:


Downloads from Sourceforge are available as well.

Feedback, issues, ideas?

To get in touch, use the following channels:

Back to top