Skip to content

mathpere/grails-hibernate-search-plugin

BranchesTags

Repository files navigation

Grails Hibernate Search Plugin

This plugin aims to integrate Hibernate Search features to Grails in very few steps.

Getting started

If you don't want to start from the template project, you could start a fresh project:

And add the following to your dependencies

  compile("org.grails.plugins:hibernate-search:2.3.0")
  compile("org.grails.plugins:hibernate5:6.1.8")
  compile("org.grails.plugins:cache")
  compile("org.hibernate:hibernate-core:5.2.10.Final")
  compile("org.hibernate:hibernate-ehcache:5.2.10.Final")

Configuration

By default, the plugin stores your indexes in this directory:

 ~/.grails/${grailsVersion}/projects/${yourProjectName}/lucene-index/development/

You can override this configuration in your application.yml

hibernate:
    cache:
        use_second_level_cache: true
        use_query_cache: true
        provider_class: net.sf.ehcache.hibernate.EhCacheProvider
        region:
            factory_class: org.hibernate.cache.ehcache.SingletonEhCacheRegionFactory
    search:
        default:
        	indexBase: '/path/to/your/indexes'
            indexmanager: near-real-time
            directory_provider: filesystem

You can also define the path to your indexes with JNDI configuration as following:

hibernate:
    cache:
        use_second_level_cache: true
        use_query_cache: true
        provider_class: net.sf.ehcache.hibernate.EhCacheProvider
        region:
            factory_class: org.hibernate.cache.ehcache.SingletonEhCacheRegionFactory
    search:
        default:
            indexBaseJndiName: 'java:comp/env/luceneIndexBase'
            directory_provider: filesystem

Indexing

Mark your domain classes as indexable

Add a static lucenceIndexing closure as following:

Note: You can use properties from super class and traits with no additional configuration (since 2.0.2)

class MyDomainClass {

    String author
    String body
    Date publishedDate
    String summary
    String title
    Status status
    Double price
    Integer someInteger

    enum Status {
        DISABLED, PENDING, ENABLED
    }

    static hasMany = [categories: Category, items: Item]

    static luceneIndexing = {
        // fields
        author index: 'yes'
        body termVector: 'with_positions'
        publishedDate date: 'day'
        summary boost: 5.9
        title index: 'yes', sortable: [name: title_sort, normalizer: LowerCaseFilterFactory]
        status index: 'yes', sortable: true
        categories indexEmbedded: true
        items indexEmbedded: [depth: 2] // configure the depth indexing
        price numeric: 2, analyze: false
        someInteger index: 'yes', bridge: ['class': PaddedIntegerBridge, params: ['padding': 10]]

        // support for classBridge
        classBridge = ['class': MyClassBridge, params: [myParam: "4"]]
    }

}

This static property indicates which fields should be indexed and describes how the field has to be indexed.

Also, the plugin lets you to mark your domain classes as indexable with the Hibernate Search annotations.

@Indexed
@ClassBridge(
     impl = MyClassBridge,
     params = @Parameter( name="myParam", value="4" ) )
class MyDomainClass {

    // when using annotations, id is required to define DocumentId
    @DocumentId
    Long id

    @Field(index=Index.YES)
    String author

    @Field(index=Index.YES)
    String body

    @Field
    @DateBridge(resolution=Resolution.DAY)
    Date publishedDate

    @Field(index=Index.YES)
    String summary

    @Field(index=Index.YES)
    @Field(name="title_sort", normalizer=@Normalizer(impl=LowerCaseFilterFactory))
    @SortableField(forField="title_sort")
    String title

    @Field(index=Index.YES)
    @SortableField
    Status status

    @Field
    @NumericField( precisionStep = 2)
    Double price

    @Field(index=Index.YES)
    @FieldBridge(impl = PaddedIntegerBridge.class, params = @Parameter(name="padding", value="10"))
    Integer someInteger

    enum Status {
        DISABLED, PENDING, ENABLED
    }

    @IndexedEmbedded
    Set categories

    @IndexedEmbedded(depth = 2)
    Set items

    static hasMany = [categories: Category, items: Item]

}

Indexing the data

Create index for existing data

The plugin lets you to create index of any indexed entity as following:

MyDomainClass.search().createIndexAndWait()

This method relies on MassIndexer and can be configured like this:

MyDomainClass.search().createIndexAndWait {
   ...
   batchSizeToLoadObjects 25
   cacheMode org.hibernate.CacheMode.NORMAL
   threadsToLoadObjects 5
   ...
}

#### Manual index changes

##### Adding instances to index

```groovy

// index only updated at commit time
MyDomainClass.search().withTransaction { transaction ->
   MyDomainClass.findAll().each {
      it.search().index()
   }
}

Deleting instances from index

// index only updated at commit time
MyDomainClass.search().withTransaction { transaction ->

   MyDomainClass.get(3).search().purge()

}

To remove all entities of a given type, you could use the following purgeAll method:

// index only updated at commit time
MyDomainClass.search().withTransaction {
   MyDomainClass.search().purgeAll()
}

Rebuild index on start

Hibernate Search offers an option to rebuild the whole index using the MassIndexer API. This plugin provides a configuration which lets you to rebuild automatically your indexes on startup.

To use the default options of the MassIndexer API, simply provide this option into your runtime.groovy:

grails.plugins.hibernatesearch = {
    rebuildIndexOnStart true
}

If you need to tune the MassIndexer API, you could specify options with a closure as following:

grails.plugins.hibernatesearch = {

    rebuildIndexOnStart {
		batchSizeToLoadObjects 30
		threadsForSubsequentFetching 8 	
		threadsToLoadObjects 4
		threadsForIndexWriter 3
		cacheMode CacheMode.NORMAL
    }

}

Search

The plugin provides you dynamic method to search for indexed entities.

Retrieving the results

All indexed domain classes provides .search() method which lets you to list the results. The plugin provides a search DSL for simplifying the way you can search. Here is what it looks like with the search DSL: (See the HibernateSearchQueryBuilder class to check the available methods)

class SomeController {

   def myAction = { MyCommand command ->

      def page = [max: Math.min(params.max ? params.int('max') : 10, 50), offset: params.offset ? params.int('offset') : 0]

      def myDomainClasses = MyDomainClass.search().list {

         if ( command.dateTo ) {
            below "publishedDate", command.dateTo
         }

         if ( command.dateFrom ) {
            above "publishedDate", command.dateFrom
         }

         mustNot {
            keyword "status", Status.DISABLED
         }

         if ( command.keyword ) {
            should {
               command.keyword.tokenize().each { keyword ->

                  def wild = keyword.toLowerCase() + '*'

                  wildcard "author", wild
                  wildcard "body", wild
                  wildcard "summary", wild
                  wildcard "title", wild
                  wildcard "categories.name", wild
               }
            }
         }

         sort "publishedDate", "asc"

         maxResults page.max

         offset page.offset
      }

      [myDomainClasses: myDomainClasses]
   }
}

Mixing with criteria query

Criteria criteria = fullTextSession.createCriteria( clazz ).createAlias("session", "session").add(Restrictions.eq("session.id", 115L));

  def myDomainClasses = MyDomainClass.search().list {

    criteria {
       setFetchMode("authors", FetchMode.JOIN)
    }

    fuzzy "description", "mi search"
  }

Performing SimpleQueryString searches

See Hibernate Search Simple Query Strings for more details on the actual query string. You can implement any other queries alongside a simple query string search.

Simple search on 1 field

// Search for "war and peace or harmony" in the title field
def myDomainClasses = MyDomainClass.search().list {
	simpleQueryString 'war + (peace | harmony)', 'title'
}

Simple search on multiple fields

// Search for "war and peace or harmony" in the title and description field
def myDomainClasses = MyDomainClass.search().list {
	simpleQueryString 'war + (peace | harmony)', 'title', 'description
}

Simple search on field setting AND as the default operator

// Search for "war and peace" in the title field
def myDomainClasses = MyDomainClass.search().list {
	simpleQueryString 'war peace', [withAndAsDefaultOperator: true], 'title'
}

Simple search on multiple fields setting the boost for each field

// Search for "war and peace" in the title field and description field with boosts applied
def myDomainClasses = MyDomainClass.search().list {
	simpleQueryString 'war + (peace | harmony)', ['title':2.0, 'description':0.5]
}

Sorting the results

sort() method accepts an optional second parameter to specify the sort order: "asc"/"desc". Default is "asc".

Fields used for sorting can be analyzed, but must not be tokenized, so you should rather use normalizers on those fields.

If you try to sort on an indexed field which has not been marked as "sortable" you will either get warnings or full errors. Therefore it is important to mark any indexed fields as sortable, and as sortable fields cannot be indexed with tokenizer analyzers you should also define a normalizer to be used (see the section on Normalizers on how to define them).

MyDomainClass.search().list {
   ...
   sort "publishedDate", "asc"
   ...  
}

If for some reasons, you want to sort results with a property which doesn't exist in your domain class, you should specify the sort type with a third parameter (default is String). You have three ways to achieve this:

By Specifying the type (could be Integer, String, Date, Double, Float, Long, Bigdecimal):

MyDomainClass.search().list {
   ...
   sort "my_special_field", "asc", Integer
   ...
}

By Specifying directly its sort field (Lucene):

def items = Item.search().list {
  ...
  sort "my_special_field", "asc", org.apache.lucene.search.SortField.Type.STRING_VAL
  ...
}

By specifying its sort field with string:

def items = Item.search().list {
  ...
  sort "my_special_field", "asc", "string_val"
  ...
}

Counting the results

You can also retrieve the number of results by using 'count' method:

def myDomainClasses = MyDomainClass.search().count {
 ...
}

Additional features

Support for ignoreAnalyzer(), ignoreFieldBridge() and boostedTo() functions

When searching for data, you may want to not use the field bridge or the analyzer. All methods (below, above, between, keyword, fuzzy) accept an optional map parameter to support this:

MyDomainClass.search().list {

   keyword "status", Status.DISABLED, [ignoreAnalyzer: true]

   wildcard "description", "hellow*", [ignoreFieldBridge: true, boostedTo: 1.5f]

}

Fuzzy search

On fuzzy search, you can add an optional parameter to specify the max distance

See Hibernate Fuzzy Search

MyDomainClass.search().list {

   keyword "status", Status.DISABLED, [ignoreAnalyzer: true]

   fuzzy "description", "hellow", [ignoreFieldBridge: true, maxDistance: 2]

}

Support for projections

Hibernate Search lets you to return only a subset of properties rather than the whole domain object. It makes it possible to avoid to query the database. This plugin supports this feature:

def myDomainClasses = MyDomainClass.search().list {

    projection "author", "body"

}

myDomainClasses.each { result ->

    def author = result[0]
    def body  = result[1]

    ...
}

Don't forget to store the properties into the index as following:

class MyDomainClass {

    [...]

    static luceneIndexing = {
        author index: 'yes', store: 'yes'
        body index: 'yes', store: 'yes'
    }
}

Analysis

Define named analyzers

Named analyzers are global and can be defined within runtime.groovy as following:

import org.apache.solr.analysis.StandardTokenizerFactory
import org.apache.solr.analysis.LowerCaseFilterFactory
import org.apache.solr.analysis.NGramFilterFactory

...

grails.plugins.hibernatesearch = {

    analyzer( name: 'ngram', tokenizer: StandardTokenizerFactory ) {
        filter LowerCaseFilterFactory
        filter factory: NGramFilterFactory, params: [minGramSize: 3, maxGramSize: 3]
    }

}

This configuration is strictly equivalent to this annotation configuration:

@AnalyzerDef(name = "ngram", tokenizer = @TokenizerDef(factory = StandardTokenizerFactory.class),
  filters = {
    @TokenFilterDef(factory = LowerCaseFilterFactory.class),
    @TokenFilterDef(factory = NGramFilterFactory.class,
      params = {
        @Parameter(name = "minGramSize",value = "3"),
        @Parameter(name = "maxGramSize",value = "3")
     })
})
public class Address {
...
}

Use named analyzers

Set the analyzer at the entity level: all fields will be indexed with the analyzer

class MyDomainClass {

    String author
    String body
    ...

    static luceneIndexing = {
        analyzer = 'ngram'
        author index: 'yes'
        body index: 'yes'
    }

}

Or set the analyzer at the field level:

class MyDomainClass {

    String author
    String body
    ...

    static luceneIndexing = {
        author index: 'yes'
        body index: 'yes', analyzer: 'ngram'
        other index: 'yes', analyzer: new MyFilter()
    }

}

Get scoped analyzer for given entity

The plugin lets you ro retrieve the scoped analyzer for a given analyzer with the search() method:

def parser = new org.apache.lucene.queryParser.QueryParser (
    "title", Song.search().getAnalyzer() )

Normalizer

Normalizers are analyzers without tokenization and are important for indexed fields which you want to sort, see Hibernate Search Normalizer for more information.

Define named normalizers

Named normalizers are global and can be defined within runtime.groovy as following:

import org.apache.lucene.analysis.core.LowerCaseFilterFactory
import org.apache.lucene.analysis.miscellaneous.ASCIIFoldingFilterFactory

...

grails.plugins.hibernatesearch = {

    normalizer(name: 'lowercase') {
        filter ASCIIFoldingFilterFactory
        filter LowerCaseFilterFactory
    }

}

This configuration is strictly equivalent to this annotation configuration:

@NormalizerDef(name = "lowercase",
  filters = {
    @TokenFilterDef(factory = ASCIIFoldingFilterFactory.class),
    @TokenFilterDef(factory = LowerCaseFilterFactory.class)
})
public class Address {
...
}

Use named normalizer

Set the normalizer at the field level

class MyDomainClass {

    String author
    String body
    ...

    static luceneIndexing = {
        author index: 'yes', sortable: [name: author_sort, normalizer: 'lowercase']
        body index: 'yes', sortable: [name: author_sort, normalizer: LowerCaseFilterFactory]
    }

}

Filters

In Hibernate Search 5.9.x the Filter class is completely removed and filters must now be applied as Full-Text Filters which are passed Querys rather than Filters.

Define named filters

Named filters are global and MUST be defined within runtime.groovy as following:

...

grails.plugins.hibernatesearch = {

    // cf official doc https://docs.jboss.org/hibernate/stable/search/reference/en-US/html_single/#query-filter-fulltext
    // Example 116. Defining and implementing a Filter
    fullTextFilter name: "bestDriver", impl: BestDriversFilter

    // cf official doc https://docs.jboss.org/hibernate/stable/search/reference/en-US/html_single/#query-filter-fulltext
    // Example 118. Using parameters in the actual filter implementation    
    fullTextFilter name: "security", impl: SecurityFilterFactory, cache: "instance_only"

}

If they are not defined in runtime.groovy they will not be available for querying.

Filter query results

Filter query results looks like this:

MyDomainClass.search().list {

// without params:
MyDomainClass.search().list {
  ...
  filter "bestDriver"
  ...
}

// with params:
MyDomainClass.search().list {
  ...
   filter name: "security", params: [ level: 4 ]
  ...
}

Options

grails.plugins.hibernatesearch = {
	rebuildIndexOnStart false // see related section above
	throwOnEmptyQuery false // throw or not exception when Hibernate Search raises an EmptyQueryException
	fullTextFilter /* ... */ // see related section above
}

Notes

Updating from 2.2 to 2.3

There is a signification change between 2.2 and 2.3.

Updating filters

Filters must now be defined in the runtime.groovy in advance and then added to a query as filter definitions which will define fullTextFilters. This is due to the deprecation of the filter class from Hibernate Search.

runtime.groovy vs application.groovy

In Grails 3 the application.groovy file is loaded when the Grails CLI is started, therefore certain logic and requirements on dependencies will fall over when defined in the application.groovy file.

The solution is to define a runtime.groovy file and move the logic into this file, this also helps to provide a nice divide on what logic is required when running the application and as config is now provided in the application.yml file it should result in only needing to define a runtime.groovy file and not the application.groovy file.

We therefore advise all hibernatesearch closure config to be defined in the runtime.groovy file.

runtime.groovy is run along with application.groovy when the application starts up, it is also packaged and run by a WAR.

IDE Integration

Unfortunately IDEs will not recognise the search() method as it is added dynamically. One messy but possible way to get around this and gain access to the DSL inside the IDE is to add an extra static method to your class. This is not ideal but it may make your programming easier.

class DomainClass {

    ...

    static List<DomainClass> hibernateSearchList(@DelegatesTo(HibernateSearchApi) Closure closure){
        DomainClass.search().list(closure)
    }
    
    static int hibernateSearchCount(@DelegatesTo(HibernateSearchApi) Closure closure){
        DomainClass.search().count(closure)
    }
}

SessionFactory failures during startup

During the SessionFactory build process any exceptions which occur during the HibernateSearch boot sequence are silently wrapped and hidden inside the futures. This means there will be a particularly helpful exception thrown :

Exception encountered during context initialization - cancelling refresh attempt: org.springframework.beans.factory.UnsatisfiedDependencyException: Error creating bean with name 'methodValidationPostProcessor' defined in class path resource [org/springframework/boot/autoconfigure/validation/ValidationAutoConfiguration.class]: Unsatisfied dependency expressed through method 'methodValidationPostProcessor' parameter 0; nested exception is org.springframework.beans.factory.BeanCreationException: Error creating bean with name 'hibernateDatastoreServiceRegistry': Cannot resolve reference to bean 'hibernateDatastore' while setting constructor argument; nested exception is org.springframework.beans.factory.BeanCreationException: Error creating bean with name 'hibernateDatastore': Bean instantiation via constructor failed; nested exception is org.springframework.beans.BeanInstantiationException: Failed to instantiate [org.grails.orm.hibernate.HibernateDatastore]: Constructor threw exception; nested exception is java.lang.NullPointerException

which will stacktrace down to :

Caused by: java.lang.NullPointerException: null
	at org.grails.orm.hibernate.cfg.HibernateMappingContextConfiguration$2.sessionFactoryClosed(HibernateMappingContextConfiguration.java:266)
	at org.hibernate.internal.SessionFactoryObserverChain.sessionFactoryClosed(SessionFactoryObserverChain.java:61)
	at org.hibernate.internal.SessionFactoryImpl.close(SessionFactoryImpl.java:756)
	at org.hibernate.search.hcore.impl.HibernateSearchSessionFactoryObserver.boot(HibernateSearchSessionFactoryObserver.java:134)
	at org.hibernate.search.hcore.impl.HibernateSearchSessionFactoryObserver.sessionFactoryCreated(HibernateSearchSessionFactoryObserver.java:79)
	at org.hibernate.internal.SessionFactoryObserverChain.sessionFactoryCreated(SessionFactoryObserverChain.java:35)
	at org.hibernate.internal.SessionFactoryImpl.<init>(SessionFactoryImpl.java:366)
	at org.hibernate.boot.internal.SessionFactoryBuilderImpl.build(SessionFactoryBuilderImpl.java:452)
	at org.hibernate.cfg.Configuration.buildSessionFactory(Configuration.java:710)
	at org.grails.orm.hibernate.cfg.HibernateMappingContextConfiguration.buildSessionFactory(HibernateMappingContextConfiguration.java:274)
	at grails.plugins.hibernate.search.context.HibernateSearchMappingContextConfiguration.buildSessionFactory(HibernateSearchMappingContextConfiguration.java:357)
	at org.grails.orm.hibernate.connections.HibernateConnectionSourceFactory.create(HibernateConnectionSourceFactory.java:86)
	at org.grails.orm.hibernate.connections.AbstractHibernateConnectionSourceFactory.create(AbstractHibernateConnectionSourceFactory.java:39)
	at org.grails.orm.hibernate.connections.AbstractHibernateConnectionSourceFactory.create(AbstractHibernateConnectionSourceFactory.java:23)
	at org.grails.datastore.mapping.core.connections.AbstractConnectionSourceFactory.create(AbstractConnectionSourceFactory.java:64)
	at org.grails.datastore.mapping.core.connections.AbstractConnectionSourceFactory.create(AbstractConnectionSourceFactory.java:52)
	at org.grails.datastore.mapping.core.connections.ConnectionSourcesInitializer.create(ConnectionSourcesInitializer.groovy:24)
	at org.grails.orm.hibernate.HibernateDatastore.<init>(HibernateDatastore.java:196)
	at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
	at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
	at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
	at java.lang.reflect.Constructor.newInstance(Constructor.java:423)
	at org.springsource.loaded.ri.ReflectiveInterceptor.jlrConstructorNewInstance(ReflectiveInterceptor.java:1076)
	at org.springframework.beans.BeanUtils.instantiateClass(BeanUtils.java:142)
	... 51 common frames omitted

The actual exception stack is not at all helpful as whilst the actual failure point is in:

org.hibernate.search.hcore.impl.HibernateSearchSessionFactoryObserver.boot(HibernateSearchSessionFactoryObserver.java:134)

It is masked inside the catch statement at line 127 inside the class, as the finally clause is what results in the above exception stacktrace.

public class HibernateSearchSessionFactoryObserver implements SessionFactoryObserver {
    // ...
    
    private synchronized void boot(SessionFactory factory) {
        try{ 
            // ...
        }
        catch (Throwable t) {
            extendedSearchIntegratorFuture.completeExceptionally( t );
            // This will make the SessionFactory abort and close itself
            throw t;
        }finally {
            if ( failedBoot ) {
                factory.close();
            }
        }
    }
    
    // ...
}

Therefore if you get the above exceptions then drop a debug point at line 130 and then start with a debugger running. The debug point will give you the helpful exception as to why the boot has failed.

Examples

A sample project is available at this repository URL https://github.com/lgrignon/grails3-quick-start

It contains several branches for each version of this plugin

Change log

v2.3

  • Grails 3.3.x
  • GORM 6.1
  • Hibernate 5.2.10
  • Hibernate Search 5.9.1
  • Add sortable field
  • Add SimpleQueryString

v2.2

  • Grails 3.3.x
  • GORM 6.1
  • Hibernate 5.2.9
  • Hibernate Search 5.7

v2.1.2

  • Supports hibernate.configClass if any
  • Removed dependencies to info.app.grailsVersion, info.app.name

v2.1

  • Grails 3.2.x
  • GORM 6
  • Hibernate 5.2.9
  • Hibernate Search 5.7

v2.0.2

Support for indexing trait properties

v2.0.1

Support for indexing inherited properties

v2.0

  • Grails 3.1.x
  • GORM 5
  • Hibernate 5.1.1
  • Hibernate Search 5.5.4

v1.x

  • Grails 2.x
  • Hibernate 4

Authors

Mathieu Perez

Julie Ingignoli

Louis Grignon

Development / Contribution

Install with:

gradlew clean publishToMavenLocal

Publish with:

gradlew clean bintrayUpload --stacktrace -PbintrayUser=... -PbintrayKey=...

License

Licensed under the Apache License, Version 2.0: https://www.apache.org/licenses/LICENSE-2.0

About

Integrates Hibernate Search features to Grails

grails.org/plugin/hibernate-search

Resources

Readme

License

Apache-2.0 license
Activity

Stars

21 stars

Watchers

3 watching

Forks

18 forks
Report repository

Releases 6

Hibernate search 5.9.1 and sortable Latest
May 26, 2018
+ 5 releases

Packages

No packages published

Contributors 5

Languages