Preface

Hibernate OGM is a persistence engine providing Java Persistence (JPA) support for NoSQL datastores. It reuses Hibernate ORM’s object life cycle management and (de)hydration engine but persists entities into a NoSQL store (key/value, document, column-oriented, etc) instead of a relational database.

It allows using the Java Persistence Query Language (JPQL) as an interface to querying stored data, in addition to using native queries of the specific NoSQL database.

The project is now fairly mature when it comes to the storage strategies, and the feature set is sufficient to be used in your projects. We do have however much bigger ambitions than a simple object mapper. Many things are on the roadmap (more NoSQL, query, denormalization engine, etc). If you feel a feature is missing, report it to us. If you want to contribute, even better!

Hibernate OGM is released under the LGPL open source license.

The future of this project is being shaped by the requests from our users. Please give us feedback on

  • what you like

  • what you don’t like

  • what is confusing

  • what you are missing as a feature

Check How to contribute on how to contact us.

We worked hard on this documentation but we know it is far from perfect. If you find something confusing or feel that an explanation is missing, please let us know. Getting in touch is easy: see contacting the developer community.

Goals

Hibernate OGM:

  • offers a familiar programming paradigm (JPA) to deal with NoSQL stores

  • moves model denormalization from a manual imperative work to a declarative approach handled by the engine

  • encourages new data usage patterns and NoSQL exploration in more "traditional" enterprises

  • helps scale existing applications with a NoSQL front end to a traditional database

NoSQL can be very disconcerting as it is composed of many disparate solutions with different benefits and drawbacks. NoSQL databases can be loosely classified in four families:

  • graph oriented databases

  • key/value stores: essentially Maps but with different behaviors and ideas behind various products (data grids, persistent with strong or eventual consistency, etc)

  • document based datastores: maps which contain semi-structured documents (think JSON)

  • column based datastores

nosql
Figure 1. Various NoSQL families

Each have different benefits and drawbacks and one solution might fit a use case better than an other. However access patterns and APIs are different from one product to the other.

Hibernate OGM is not expected to be the Rosetta stone used to interact with all NoSQL solution in all use cases. But for people modeling their data as a domain model, it provides distinctive advantages over raw APIs and has the benefit of providing an API and semantic known to Java developers. Reusing the same programmatic model and trying different (No)SQL engines will hopefully help people to explore alternative datastores.

Hibernate OGM also aims at helping people scale traditional relational databases by providing a NoSQL front-end and keeping the same JPA APIs and domain model. It could for example help to migrate a selection of your model from an RDBMS to a particular NoSQL solution which better fits the typical use case.

What we have today

Today, Hibernate OGM does not support all of these goals. Here is a list of what we have:

  • store data in key/value stores (Infinispan and Ehcache)

  • store data in document stores (MongoDB and CouchDB - the latter in preview)

  • store data in graph databases (Neo4J)

  • Create, Read, Update and Delete operations (CRUD) for entities

  • polymorphic entities (support for superclasses, subclasses etc).

  • embeddable objects (aka components)

  • support for basic types (numbers, String, URL, Date, enums, etc)

  • support for associations

  • support for collections (Set, List, Map, etc)

  • support for JPQL queries (not arbitrary joins though)

  • support for mapping native queries results to managed entities

  • support for Hibernate Search’s full-text queries

  • and generally, support for JPA and native Hibernate ORM API support

In short, a perfectly capable Object Mapper for multiple popular NoSQL datastores.

Experimental features

As Hibernate OGM is a rather young project, some parts of it may be marked as experimental. This may affect specific APIs or SPIs (e.g. the case for the SchemaInitializer SPI contract at the moment), entire dialects (this is the case for the CouchDB dialect at the moment) or deliverables.

Experimental APIs/SPIs are marked via the @Experimental annotation. Experimental dialects make that fact apparent through their datastore name (e.g. "COUCHDB_EXPERIMENTAL") and experimental deliverables use the "experimental" artifact classifier.

If a certain part is marked as experimental it may undergo backwards-incompatible changes in future releases. E.g. API/SPI methods may be altered, so that code using them needs to be adapted as well. For experimental dialects the persistent format of data may be changed, so that a future version of such dialect may not be able to read back data written by previous versions. A manual update of the affected data may be thus required. Experimental deliverables should be used with special care, as they are work in progress. You should use them for testing but not production use cases.

But most of our dialects are mature, so don’t worry ;)

Use cases

Here are a few areas where Hibernate OGM can be beneficial:

  • need to scale your datastore up and down rapidly (via the underlying NoSQL datastore capability)

  • keep your domain model independent of the underlying datastore technology (RDBMS, Infinispan, NoSQL)

  • explore the best tool for the use case

  • use a familiar JPA front end to your datastore

  • use Hibernate Search full-text search / text analysis capabilities and store the data set in an scalable datastore

These are a few ideas and the list will grow as we add more capabilities to Hibernate OGM.

1. How to get help and contribute on Hibernate OGM

Hibernate OGM is a young project. Join and help us shape it!

1.1. How to get help

First of all, make sure to read this reference documentation. This is the most comprehensive formal source of information. Of course, it is not perfect: feel free to come and ask for help, comment or propose improvements in our Hibernate OGM forum.

You can also:

  • open bug reports in JIRA

  • propose improvements on the development mailing list

  • join us on IRC to discuss developments and improvements (#hibernate-dev on freenode.net; you need to be registered on freenode: the room does not accept "anonymous" users).

1.2. How to contribute

Welcome!

There are many ways to contribute:

  • report bugs in JIRA

  • give feedback in the forum, IRC or the development mailing list

  • improve the documentation

  • fix bugs or contribute new features

  • propose and code a datastore dialect for your favorite NoSQL engine

Hibernate OGM’s code is available on GitHub at https://github.com/hibernate/hibernate-ogm.

1.2.1. How to build Hibernate OGM

Hibernate OGM uses Git and Maven 3, make sure to have both installed on your system.

Clone the git repository from GitHub:

#get the sources
git clone https://github.com/hibernate/hibernate-ogm
cd hibernate-ogm

Run maven

#build project
mvn clean install -s settings-example.xml

Note that Hibernate OGM uses artifacts from the Maven repository hosted by JBoss. Make sure to either use the -s settings-example.xml option or adjust your ~/.m2/settings.xml according to the descriptions available on this jboss.org wiki page.

These settings are required for development of Hibernate OGM but should not be needed to use it.

To skip building the documentation, set the skipDocs property to true:

mvn clean install -DskipDocs=true -s settings-example.xml

If you just want to build the documentation only, run it from the hibernate-ogm-documentation/manual subdirectory.

1.2.2. How to contribute code effectively

The best way to share code is to fork the Hibernate OGM repository on GitHub, create a branch and open a pull request when you are ready. Make sure to rebase your pull request on the latest version of the master branch before offering it.

Here are a couple of approaches the team follows:

  • We do small independent commits for each code change. In particular, we do not mix stylistic code changes (import, typos, etc) and new features in the same commit.

  • Commit messages follow this convention: the JIRA issue number, a short commit summary, an empty line, a longer description if needed. Make sure to limit line length to 80 characters, even at this day and age it makes for more readable commit comments.

OGM-123 Summary of commit operation

Optional details on the commit
and a longer description can be
added here.
  • A pull request can contain several commits but should be self contained: include the implementation, its unit tests, its documentation and javadoc changes if needed.

  • All commits are proposed via pull requests and reviewed by another member of the team before being pushed to the reference repository. That’s right, we never commit directly upstream without code review.

1.3. How to build support for a data store

Advanced section

This is an advanced subject, feel free to skip this section if you are not building a data store.

Hibernate OGM supports various data stores by abstracting them with DatastoreProvider and GridDialect. The supported features vary between data stores, and dialects do not have to implement all features. Hibernate OGM implements a TCK (Technology Compatibility Kit) to verify interoperability and features of the dialect. Hibernate OGM supports a variety of document- and key-value-stores and ships with some abstraction and utility classes for document- and key-value-stores (like KeyValueStoreProperties and DocumentStoreProperties).

1.3.1. DataStore providers

Supporting a data store usually begins with a DatastoreProvider. Providers can implement a lifecycle (start, stop) to initialize, configure and shutdown resources. Taking a look at existing data store support such as MongoDB (see org.hibernate.ogm.datastore.mongodb.impl.MongoDBDatastoreProvider) is a good idea to get an impression of how to boot the data store support. Providers are seen as services, they can implement various service interfaces to activate certain features (see the org.hibernate.service.spi package for details).

A common issue to face then implementing new data stores is transactionality. Some data stores provide transactional support that can be used in the context of Hibernate OGM wrapped by JTA. If your data store does not support transactions, you can enable transaction emulation within the DatastoreProvider.

Features of a DatastoreProvider:

  • Resource lifecycle

  • Managing connection resources

  • Configuration

  • Access to query parsers

  • Define/Validate a schema

1.3.2. Dialects

A data store can have one or more dialects. Dialects describe the style how data is mapped to a particular data store. NoSQL data stores imply a certain nature, how to map data. Document-oriented data stores encourage an entity-as-document pattern where embedded data structures could be stored within the document itself. Key-value data stores allow different approaches, e.g. storing an entity as JSON document or event storing individual key-value pairs that map the entity within a hash table data structure. Hibernate OGM allows multiple dialects per data store and users may choose the most appropriate one.

The most basic support is provided by implementing the GridDialect interface. Implementing that interface is mandatory to support a specific data store.

A GridDialect usually supports:

  • Create/Read/Update/Delete for entities

  • Create/Read/Update/Delete for associations

  • Id/Sequence generator

  • Provides locking strategies

A dialect may optionally implement one or more additional facet interfaces to provide a broader support for certain features:

  • QueryableGridDialect

  • BatchableGridDialect

  • IdentityColumnAwareGridDialect

  • OptimisticLockingAwareGridDialect

  • MultigetGridDialect

Features of a QueryableGridDialect

  • Query execution

  • Support for native queries

Features of a BatchableGridDialect

  • Operation queueing

  • Execution of queued Create/Update/Delete as a batch

Features of a IdentityColumnAwareGridDialect

  • Supports the generation of identity values upon data insertion

Features of an OptimisticLockingAwareGridDialect

  • Finding and altering versioned records in an atomic fashion

Features of a MultigetGridDialect

  • Retrieve multiple tuples within one operation

Before starting make a clear plan of how you think entities, relations and nested structures are best represented in the NoSQL store you plan to implement. It helps to have a clear picture about that, and this will require some experience with the NoSQL database you plan to support.

Start with a small feature set to get a feeling for Hibernate OGM, for example aim at implementing CRUD operations only and ignore relations and queries. You can always extend the features as you proceed.

Starting from or studying existing dialects is also an interesting strategy. It can be intimidating with complex dialects though.

Hibernate OGM is not opinionated by which means data is stored/loaded for a particular data store, but the particular dialect is. Hibernate OGM strives for the most natural mapping style. The idea is to facilitate integration with other applications of that database by sticking to established patterns and idioms of that store.

1.3.3. Entities

Entities are seen by a dialect as Tuple. A Tuple contains:

  • a snapshot (that’s the view of the data as loaded from your database),

  • a set of key-value pairs that carry the actual data,

  • and a list of operations to apply onto the original snapshot. Tuple keys use dot-path property identifiers to indicate nesting. That comes handy when working with document stores because you can build a document structure based on that details.

1.3.4. Associations

Most NoSQL data stores have no built-in support for associations between entities (unless you’re using a graph database).

Hibernate OGM simulates associations for datastore with no support by storing the navigational information to go from a given entity to its (list of) associated entity. This of it as query materialisation. This navigational information data can be stored within the entity itself or externally (as own documents or relation items).

1.3.5. Configuration

Hibernate OGM can read its configuration properties from various sources. Most common configuration sources are:

  • hibernate.properties file

  • persistence.xml file

  • environment variables override or integrate properties set in the above configuration files

  • annotation configuration (entity classes)

  • programmatic configuration

The org.hibernate.ogm.options package provides the configuration infrastructure.

You might want to look at MongoDBConfiguration or InfinispanConfiguration to get an idea how configuration works. Configuration is usually read when starting a data store provider or while operating. A good example of accessing configuration during runtime is the association storage option, where users can define, how to store a particular association (within the entity or as a separate collection/key/document/node).

The configuration and options context infrastructure allows to support data store-specific options such as ReadPreference for MongoDB or TTL for Redis.

Programmatic configuration

Data store support can implement programmatic configuration. The configuration splits into three parts:

  • Global configuration

  • Entity configuration

  • Property configuration

Programmatic configuration consists of two parts: configuration interfaces (see org.hibernate.ogm.options.navigation) and partial (abstract) implementation classes. These parts are merged at runtime using ASM class generation.

1.3.6. Types

Every data store supports a unique set of data types. Some stores support floating point types and date types, others just strings. Hibernate OGM allows users to utility a variety of data types (see JPA spec) for their data models. On the other hand, that data needs to be stored within the data store and mapped back.

A dialect can provide a GridType to describe the handling of a particular data type, meaning you can specify how dates, floating point types or even byte arrays are handled. Whether they are mapped to other data types (e. g. use double for float or use base64-encoded strings for byte arrays) or wrapped within strings.

Data store-specific types can be handled the same way, check out StringAsObjectIdType for the String-mapping of MongoDB’s ObjectId type.

Type-mapping can be an exhausting task. The whole type handling is in flux and is subject to change as Hibernate OGM progresses. Ask, if you’re not sure about it.

1.3.7. Tests

Hibernate OGM brings a well suited infrastructure for tests. The test infrastructure consists of generic base classes (OgmTestCase for OGM and JpaTestCase for JPA) for tests and a test helper (see GridDialectTestHelper). That classes are used to get a different view on data than the frontend-view by the Session and the EntityManager.

It is always helpful to create a set of own test cases for different scenarios to validate the data is mapped in the way it’s intended or to verify data store-specific options such as TTL.

Another bunch of tests is called the backend TCK. That test classes test nearly all aspects of Hibernate OGM viewed from a users' perspective. Tests contain cases for simple/complex entities, associations, list- and map data types, queries using Hibernate Search, and tests for data type support.

The backend TCK is included using classpath filters, just check one of the current implementations (like RedisBackendTckHelper). When you’re developing a core module, that is included in the distribution, you will have to add your dialect to the @SkipByGridDialect annotation of some tests.

Running even 20% of the tests successfully is a great achievement. Proceed step-by-step. Large numbers of tests can fail just because of one thing that is handled differently. Don’t hesitate to ask for help.

2. Getting started with Hibernate OGM

If you are familiar with JPA, you are almost good to go. We will nevertheless walk you through the first few steps of persisting and retrieving an entity using Hibernate OGM.

Before we can start, make sure you have the following tools configured:

  • Java JDK 8

  • Maven 3.2.3 or above

Hibernate OGM is published in the Maven central repository.

Add org.hibernate.ogm:hibernate-ogm-bom:5.2.0.Beta1 to your dependency management block and org.hibernate.ogm:hibernate-ogm-infinispan:5.2.0.Beta1 to your project dependencies:

<dependencyManagement>
    <dependencies>
        <dependency>
            <groupId>org.hibernate.ogm</groupId>
            <artifactId>hibernate-ogm-bom</artifactId>
            <version>5.2.0.Beta1</version>
            <type>pom</type>
            <scope>import</scope>
        </dependency>
    </dependencies>
<dependencyManagement>

<dependencies>
    <dependency>
        <groupId>org.hibernate.ogm</groupId>
        <artifactId>hibernate-ogm-infinispan</artifactId>
    </dependency>
</dependencies>

The former is a so-called "bill of materials" POM which specifies a matching set of versions for Hibernate OGM and its dependencies. That way you never need to specify a version explicitly within your dependencies block, you will rather get the versions from the BOM automatically.

If you’re deploying your application onto WildFly 10, you don’t need to add the Hibernate OGM modules to your deployment unit but you can rather add them as modules to the application server. Refer to How to package Hibernate OGM applications for WildFly 10 to learn more.

We will use the JPA APIs in this tutorial. While Hibernate OGM depends on JPA 2.1, it is marked as provided in the Maven POM file. If you run outside a Java EE container, make sure to explicitly add the dependency:

<dependency>
    <groupId>org.hibernate.javax.persistence</groupId>
    <artifactId>hibernate-jpa-2.1-api</artifactId>
</dependency>

Let’s now map our first Hibernate OGM entity.

@Entity
public class Dog {
   @Id @GeneratedValue(strategy = GenerationType.TABLE, generator = "dog")
   @TableGenerator(
      name = "dog",
      table = "sequences",
      pkColumnName = "key",
      pkColumnValue = "dog",
      valueColumnName = "seed"
   )
   public Long getId() { return id; }
   public void setId(Long id) { this.id = id; }
   private Long id;

   public String getName() { return name; }
   public void setName(String name) { this.name = name; }
   private String name;

   @ManyToOne
   public Breed getBreed() { return breed; }
   public void setBreed(Breed breed) { this.breed = breed; }
   private Breed breed;
}

@Entity
public class Breed {

   @Id @GeneratedValue(generator = "uuid")
   @GenericGenerator(name="uuid", strategy="uuid2")
   public String getId() { return id; }
   public void setId(String id) { this.id = id; }
   private String id;

   public String getName() { return name; }
   public void setName(String name) { this.name = name; }
   private String name;
}

I lied to you, we have already mapped two entities!

If you are familiar with JPA, you can see that there is nothing specific to Hibernate OGM in our mapping.

In this tutorial, we will use JBoss Transactions for our JTA transaction manager. So let’s add the JTA API and JBoss Transactions to our POM as well.

We will also add Hibernate Search because the Infinispan dialect needs it to run JPQL queries, but we will leave the details for another time (Using Hibernate Search).

The final list of dependencies should look like this:

<dependencies>
    <!-- Hibernate OGM Infinispan module; pulls in the OGM core module -->
    <dependency>
        <groupId>org.hibernate.ogm</groupId>
        <artifactId>hibernate-ogm-infinispan</artifactId>
    </dependency>

    <1-- Optional, needed to run JPQL queries on some datastores -->
    <dependency>
        <groupId>org.hibernate.search</groupId>
        <artifactId>hibernate-search-orm</artifactId>
    </dependency>

    <!-- Standard APIs dependencies - provided in a Java EE container -->
    <dependency>
        <groupId>org.hibernate.javax.persistence</groupId>
        <artifactId>hibernate-jpa-2.1-api</artifactId>
    </dependency>
    <dependency>
        <groupId>org.jboss.spec.javax.transaction</groupId>
        <artifactId>jboss-transaction-api_1.2_spec</artifactId>
    </dependency>

    <!-- Add the Narayana Transactions Manager
         an implementation would be provided in a Java EE container,
         but this works nicely in Java SE as well -->
    <dependency>
        <groupId>org.jboss.narayana.jta</groupId>
        <artifactId>narayana-jta</artifactId>
    </dependency>
    <dependency>
        <groupId>org.jboss</groupId>
        <artifactId>jboss-transaction-spi</artifactId>
    </dependency>
</dependencies>

Next we need to define the persistence unit. Create a META-INF/persistence.xml file.

<?xml version="1.0"?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
             xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
             xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_2_0.xsd"
             version="2.0">

    <persistence-unit name="ogm-jpa-tutorial" transaction-type="JTA">
        <!-- Use the Hibernate OGM provider: configuration will be transparent -->
        <provider>org.hibernate.ogm.jpa.HibernateOgmPersistence</provider>
        <properties>
            <!-- Here you will pick which NoSQL technology to use, and configure it;
                 in this example we start a local in-memory Infinispan node. -->
            <property name="hibernate.ogm.datastore.provider" value="infinispan_embedded"/>
        </properties>
    </persistence-unit>
</persistence>

Let’s now persist a set of entities and retrieve them.

//accessing JBoss's Transaction can be done differently but this one works nicely
TransactionManager tm = com.arjuna.ats.jta.TransactionManager.transactionManager();

//build the EntityManagerFactory as you would build in in Hibernate ORM
EntityManagerFactory emf = Persistence.createEntityManagerFactory(
    "ogm-jpa-tutorial");

final Logger logger = LoggerFactory.getLogger(DogBreedRunner.class);

[..]

//Persist entities the way you are used to in plain JPA
tm.begin();
logger.infof("About to store dog and breed");
EntityManager em = emf.createEntityManager();
Breed collie = new Breed();
collie.setName("Collie");
em.persist(collie);
Dog dina = new Dog();
dina.setName("Dina");
dina.setBreed(collie);
em.persist(dina);
Long dinaId = dina.getId();
em.flush();
em.close();
tm.commit();

[..]

//Retrieve your entities the way you are used to in plain JPA
tm.begin();
logger.infof("About to retrieve dog and breed");
em = emf.createEntityManager();
dina = em.find(Dog.class, dinaId);
logger.infof("Found dog %s of breed %s", dina.getName(), dina.getBreed().getName());
em.flush();
em.close();
tm.commit();

[..]

emf.close();

A working example can be found in Hibernate OGM’s distribution under hibernate-ogm-documentation/examples/gettingstarted.

What have we seen?

  • Hibernate OGM is a JPA implementation and is used as such both for mapping and in API usage

  • It is configured as a specific JPA provider: org.hibernate.ogm.jpa.HibernateOgmPersistence

Let’s explore more in the next chapters.

Hibernate OGM might also require Hibernate Search on the classpath. This depends on the dialect or features you want to use with you project and you will find more details about it in the next chapters.

If you want to use Hibernate OGM with WildFly 10 you will need some additional configuration and you can find all the details in the How to package Hibernate OGM applications for WildFly 10 paragraph.

3. Architecture

Hibernate OGM defines an abstraction layer represented by DatastoreProvider and GridDialect to separate the OGM engine from the datastores interaction. It has successfully abstracted various key/value stores, document stores and graph databases. We are working on testing it on other NoSQL families.

In this chapter we will explore:

  • the general architecture

  • how the data is persisted in the NoSQL datastore

  • how we support JPQL queries

Let’s start with the general architecture.

3.1. General architecture

Hibernate OGM is made possible by the reuse of a few key components:

  • Hibernate ORM for JPA support

  • the NoSQL drivers to interact with the underlying datastore

  • optionally Hibernate Search for indexing and query purposes

  • optionally Infinispan’s Lucene Directory to store indexes in Infinispan itself, or in many other NoSQL using Infinispan’s write-through cachestores

  • Hibernate OGM itself

ogm architecture
Figure 2. General architecture

Hibernate OGM reuses as much as possible from the Hibernate ORM infrastructure. There is no need to rewrite a new JPA engine. The Persisters and the Loaders (two interfaces used by Hibernate ORM) have been rewritten to persist data in the NoSQL store. These implementations are the core of Hibernate OGM. We will see in How is data persisted how the data is structured.

The particularities between NoSQL stores are abstracted by the notion of a DatastoreProvider and a GridDialect.

  • DatastoreProvider abstracts how to start and maintain a connection between Hibernate OGM and the datastore.

  • GridDialect abstracts how data itself including associations are persisted.

Think of them as the JDBC layer for our NoSQL stores.

Other than these, all the Create/Read/Update/Delete (CRUD) operations are implemented by the Hibernate ORM engine (object hydration and dehydration, cascading, lifecycle etc).

As of today, we have implemented the following datastore providers:

  • a HashMap based datastore provider (for testing)

  • an Infinispan Embedded datastore provider to persist your entities in an Infinispan instance running within the same JVM

  • an Infinispan Remote datastore provider to persist your entities in Infinispan over a remote Hot Rod client

  • an Ehcache based datastore provider to persist your entities in Ehcache

  • a MongoDB based datastore provider to persist data in a MongoDB database

  • a Neo4j based datastore provider to persist data in the Neo4j graph database

  • a CouchDB based datastore provider to persist data in the CouchDB document store (experimental)

  • a Cassandra based datastore provider to persist data in Apache Cassandra (experimental)

  • a Redis based datastore provider to persist data in the Redis key/value store (experimental)

  • an Ignite based datastore provider to persist data in Apache Ignite (experimental)

To implement JPQL queries, Hibernate OGM parses the JPQL string and calls the appropriate translator functions to build a native query. Since not all NoSQL technologies support querying, when lacking we can use Hibernate Search as an external query engine.

We will discuss the subject of querying in more details in How is data queried.

Hibernate OGM best works in a JTA environment. The easiest solution is to deploy it on a Java EE container. Alternatively, you can use a standalone JTA TransactionManager. We explain how to in In a standalone JTA environment.

Let’s now see how and in which structure data is persisted in the NoSQL data store.

3.2. How is data persisted

Hibernate OGM tries to reuse as much as possible the relational model concepts, at least when they are practical and make sense in OGM’s case. For very good reasons, the relational model brought peace in the database landscape over 30 years ago. In particular, Hibernate OGM inherits the following traits:

  • abstraction between the application object model and the persistent data model

  • persist data as basic types

  • keep the notion of primary key to address an entity

  • keep the notion of foreign key to link two entities (not enforced)

If the application data model is too tightly coupled with your persistent data model, a few issues arise:

  • any change in the application object hierarchy / composition must be reflected in the persistent data

  • any change in the application object model will require a migration at the data level

  • any access to the data by another application ties both applications losing flexibility

  • any access to the data from another platform become somewhat more challenging

  • serializing entities leads to many additional problems (see note below)

Why aren’t entities serialized in the key/value entry

There are a couple of reasons why serializing the entity directly in the datastore - key/value in particular - can lead to problems:

  • When entities are pointing to other entities are you storing the whole graph? Hint: this can be quite big!

  • If doing so, how do you guarantee object identity or even consistency amongst duplicated objects? It might make sense to store the same object graph from different root objects.

  • What happens in case of class schema change? If you add or remove a property or include a superclass, you must migrate all entities in your datastore to avoid deserialization issues.

Entities are stored as tuples of values by Hibernate OGM. More specifically, each entity is conceptually represented by a Map<String,Object> where the key represents the column name (often the property name but not always) and the value represents the column value as a basic type. We favor basic types over complex ones to increase portability (across platforms and across type / class schema evolution over time). For example a URL object is stored as its String representation.

The key identifying a given entity instance is composed of:

  • the table name

  • the primary key column name(s)

  • the primary key column value(s)

data entity
Figure 3. Storing entities

The GridDialect specific to the NoSQL datastore you target is then responsible to convert this map into the most natural model:

  • for a key/value store or a data grid, we use the logical key as the key in the grid and we store the map as the value. Note that it’s an approximation and some key/value providers will use more tailored approaches.

  • for a document oriented store, the map is represented by a document and each entry in the map corresponds to a property in a document.

Associations are also stored as tuples. Hibernate OGM stores the information necessary to navigate from an entity to its associations. This is a departure from the pure relational model but it ensures that association data is reachable via key lookups based on the information contained in the entity tuple we want to navigate from. Note that this leads to some level of duplication as information has to be stored for both sides of the association.

The key in which association data are stored is composed of:

  • the table name

  • the column name(s) representing the foreign key to the entity we come from

  • the column value(s) representing the foreign key to the entity we come from

Using this approach, we favor fast read and (slightly) slower writes.

data association
Figure 4. Storing associations

Note that this approach has benefits and drawbacks:

  • it ensures that all CRUD operations are doable via key lookups

  • it favors reads over writes (for associations)

  • but it duplicates data

Again, there are specificities in how data is inherently stored in the specific NoSQL store. For example, in document oriented stores, the association information including the identifier to the associated entities can be stored in the entity owning the association. This is a more natural model for documents.

data association document
Figure 5. Storing associations in a document store

Some identifiers require to store a seed in the datastore (like sequences for examples). The seed is stored in the value whose key is composed of:

  • the table name

  • the column name representing the segment

  • the column value representing the segment

This description is how conceptually Hibernate OGM asks the datastore provider to store data. Depending on the family and even the specific datastore, the storage is optimized to be as natural as possible. In other words as you would have stored the specific structure naturally. Make sure to check the chapter dedicated to the NoSQL store you target to find the specificities.

Many NoSQL stores have no notion of schema. Likewise, the tuple stored by Hibernate OGM is not tied to a particular schema: the tuple is represented by a Map, not a typed Map specific to a given entity type. Nevertheless, JPA does describe a schema thanks to:

  • the class schema

  • the JPA physical annotations like @Table and @Column.

While tied to the application, it offers some robustness and explicit understanding when the schema is changed as the schema is right in front of the developers' eyes. This is an intermediary model between the strictly typed relational model and the totally schema-less approach pushed by some NoSQL families.

3.3. Id generation using sequences

You can use sequences with the following annotations:

  • @SequenceGenerator: it will use native sequences if available

  • @TableGenerator: it will emulate sequences storing the value in the most appropriate data structure; for example a document in MongoDB or a node in Neo4j.

Here’s some things to keep in mind when dealing with sequence generation:

  • @TableGenerator is the fallback approach used when the underlying datastore does not support native sequences generation.

  • If the datastore does not support atomic operations and does not support native sequences, Hibernate OGM will throw an exception at bootstrap and suggest alternatives.

  • The mapping of the sequence might change based on the annotation used, you should check the mapping paragraph in the documentation related to the dialect you are using.

  • The value saved in the the datastore might not be the next value in the sequence.

3.4. How is data queried

Since Hibernate OGM wants to offer all of JPA, it needs to support JPQL queries. Hibernate OGM parses the JPQL query string and extracts its meaning. From there, several options are available depending of the capabilities of the NoSQL store you target:

  • it directly delegates the native query generation to the datastore specific query translator implementation

  • it uses Hibernate Search as a query engine to execute the query

If the NoSQL datastore has some query capabilities and if the JPQL query is simple enough to be executed by the datastore, then the JPQL parser directly pushes the query generation to the NoSQL specific query translator. The query returns the list of matching entity columns or projections and Hibernate OGM returns managed entities.

Some NoSQL stores have poor query support, or none at all. In this case Hibernate OGM can use Hibernate Search as its indexing and query engine. Hibernate Search is able to index and query objects - entities - and run full-text queries. It uses the well known Apache Lucene to do this but adds a few interesting characteristics like clustering support and an object oriented abstraction including an object oriented query DSL. Let’s have a look at the architecture of Hibernate OGM when using Hibernate Search:

ogm architecture with hsearch
Figure 6. Using Hibernate Search as query engine - greyed areas are blocks already present in Hibernate OGM’s architecture

In this situation, Hibernate ORM Core pushes change events to Hibernate Search which will index entities accordingly and keep the index and the datastore in sync. The JPQL query parser delegates the query translation to the Hibernate Search query translator and executes the query on top of the Lucene indexes. Indexes can be stored in various fashions:

  • on a file system (the default in Lucene)

  • in Infinispan via the Infinispan Lucene directory implementation: the index is then distributed across several servers transparently

  • in NoSQL stores that can natively store Lucene indexes

  • in NoSQL stores that can be used as overflow to Infinispan: in this case Infinispan is used as an intermediary layer to serve the index efficiently but persists the index in another NoSQL store.

You can use Hibernate Search even if you do plan to use the NoSQL datastore query capabilities. Hibernate Search offers a few interesting options:

  • clusterability

  • full-text queries - ie Google for your entities

  • geospatial queries

  • query faceting (ie dynamic categorization of the query results by price, brand etc)

4. Configure and start Hibernate OGM

Hibernate OGM favors ease of use and convention over configuration. This makes its configuration quite simple by default.

4.1. Bootstrapping Hibernate OGM

Hibernate OGM can be used via the Hibernate native APIs (Session) or via the JPA APIs (EntityManager). Depending on your choice, the bootstrapping strategy is slightly different.

4.1.1. Using JPA

If you use JPA as your primary API, the configuration is extremely simple. Hibernate OGM is seen as a persistence provider which you need to configure in your persistence.xml. That’s it! The provider name is org.hibernate.ogm.jpa.HibernateOgmPersistence.

Example 1. persistence.xml file
<?xml version="1.0"?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
             xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
             xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_2_0.xsd"
             version="2.0">

    <persistence-unit name="org.hibernate.ogm.tutorial.jpa" transaction-type="JTA">
        <!-- Use Hibernate OGM provider: configuration will be transparent -->
        <provider>org.hibernate.ogm.jpa.HibernateOgmPersistence</provider>
        <properties>
            <property name="hibernate.transaction.jta.platform"
                      value="JBossTS" />
            <property name="hibernate.ogm.datastore.provider"
                      value="infinispan_embedded" />
        </properties>
    </persistence-unit>
</persistence>

There are a couple of things to notice:

  • there is no JDBC dialect setting

  • there is no JDBC setting except sometimes a jta-data-source (check In a Java EE container for more info)

  • most NoSQL databases do not require a schema, in which case schema generation options (hbm2ddl) do not apply

  • if you use JTA (which we recommend), you will need to set the JTA platform

You also need to configure which NoSQL datastore you want to use and how to connect to it. We will detail how to do that later in NoSQL datastores.

In this case, we have used the default settings for Infinispan: this will start a local, in-memory Infinispan instance which is useful for testing but the stored data will be lost on shutdown. You might think of this configuration as similar to storing your data in an hashmap, but you could of course change the Infinispan configuration to enable clustering (for both scalability and failover) and to enable permanent persistence strategies.

From there, simply bootstrap JPA the way you are used to with Hibernate ORM:

  • via Persistence.createEntityManagerFactory

  • by injecting the EntityManager / EntityManagerFactory in a Java EE container

  • by using your favorite injection framework (CDI - Weld, Spring, Guice)

Note that what you’re starting is not an exotic new JPA implementation but is in all effects an instance of Hibernate ORM, although using some alternative internal components to deal with the NoSQL stores. This means that any framework and tool integrating with Hibernate ORM can integrate with Hibernate OGM - of course as long as it’s not making assumptions such as that a JDBC datasource will be used.

4.1.2. Using Hibernate ORM native APIs

If you want to bootstrap Hibernate OGM using the native Hibernate APIs, use the new bootstrap API from Hibernate ORM 5. By setting OgmProperties.ENABLED to true, the Hibernate OGM components will be activated. Note that unwrapping into OgmSessionFactoryBuilder is not strictly needed, but it will allow you to set Hibernate OGM specific options in the future and also gives you a reference to OgmSessionFactory instead of SessionFactory.

Example 2. Bootstrap Hibernate OGM with Hibernate ORM native APIs
StandardServiceRegistry registry = new StandardServiceRegistryBuilder()
    .applySetting( OgmProperties.ENABLED, true )
    //assuming you are using JTA in a non container environment
    .applySetting( AvailableSettings.TRANSACTION_COORDINATOR_STRATEGY, "jta" )
    //assuming JBoss TransactionManager in standalone mode
    .applySetting( AvailableSettings.JTA_PLATFORM, "JBossTS" )
    //assuming Infinispan as the backend, using the default settings
    .applySetting( OgmProperties.DATASTORE_PROVIDER, InfinispanEmbedded.DATASTORE_PROVIDER_NAME );
    .build();

//build the SessionFactory
OgmSessionFactory sessionFactory = new MetadataSources( registry )
    .addAnnotatedClass( Order.class )
    .addAnnotatedClass( Item.class )
    .buildMetadata()
    .getSessionFactoryBuilder()
    .unwrap( OgmSessionFactoryBuilder.class )
    .build();

There are a couple of things to notice:

  • there is no DDL schema generation options (hbm2ddl) as Infinispan does not require schemas when running in embedded mode

  • you need to set the right transaction strategy and the right transaction manager lookup strategy if you use a JTA based transaction strategy (see Environments)

You also need to configure which NoSQL datastore you want to use and how to connect to it. We will detail how to do that later in NoSQL datastores. In this case, we have used the defaults settings for Infinispan.

4.2. Environments

Hibernate OGM runs in various environments: it should work pretty much in all environments in which Hibernate ORM runs. There are however some selected environments in which it was tested more thoroughly than others. The current version is being tested regularly in Java SE (without a container) and within the WildFly 10 application server; at time of writing this there’s no known reason for it to not work in different containers as long as you remember that it requires a specific version of Hibernate ORM: some containers might package a conflicting version.

4.2.1. In a Java EE container

You don’t have to do much in this case. You need three specific settings:

  • the transaction coordinator type

  • the JTA platform

  • a JTA datasource

If you use JPA, simply set the transaction-type to JTA and the transaction factory will be set for you.

If you use Hibernate ORM native APIs only, then set hibernate.transaction.coordinator_class to "jta".

Set the JTA platform to the right Java EE container. The property is hibernate.transaction.jta.platform and must contain the fully qualified class name of the lookup implementation. The list of available values are listed in Hibernate ORM’s configuration section. For example in WildFly 10 you would pick JBossAS, although in WildFly these settings are automatically injected so you could skip this.

In your persistence.xml you usually need to define an existing datasource. This is not needed by Hibernate OGM: it will ignore the datasource, but JPA specification mandates the setting.

Example 3. persistence.xml file
<?xml version="1.0"?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
             xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
             xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_2_0.xsd"
             version="2.0">

    <persistence-unit name="org.hibernate.ogm.tutorial.jpa" transaction-type="JTA">
        <!-- Use Hibernate OGM provider: configuration will be transparent -->
        <provider>org.hibernate.ogm.jpa.HibernateOgmPersistence</provider>
        <jta-data-source>java:/DefaultDS</jta-data-source>
        <properties>
            <property name="hibernate.transaction.jta.platform" value="JBossAS" />
            <property name="hibernate.ogm.datastore.provider" value="infinispan_embedded" />
        </properties>
    </persistence-unit>
</persistence>

java:DefaultDS will work for out of the box WildFly deployments.

4.2.2. In a standalone JTA environment

There is a set of common misconceptions in the Java community about JTA:

  • JTA is hard to use

  • JTA is only needed when you need transactions spanning several databases

  • JTA works in Java EE only

  • JTA is slower than "simple" transactions

None of these are true: let me show you how to use the Narayana Transactions Manager in a standalone environment with Hibernate OGM.

In Hibernate OGM, make sure to set the following properties:

  • transaction-type to JTA in your persistence.xml if you use JPA

  • or hibernate.transaction.coordinator_class to "jta" if you use StandardServiceRegistryBuilder/OgmConfiguration to bootstrap Hibernate OGM.

  • hibernate.transaction.jta.platform to JBossTS in both cases.

Add the Narayana Transactions Manager to your classpath. If you use maven, it should look like this:

Example 4. Narayana Transactions Manager dependency declaration
<dependency>
    <groupId>org.jboss.narayana.jta</groupId>
    <artifactId>narayana-jta</artifactId>
    <version>5.4.0.Final</version>
</dependency>

The next step is you get access to the transaction manager. The easiest solution is to do as the following example:

TransactionManager transactionManager =
   com.arjuna.ats.jta.TransactionManager.transactionmanager();

Then use the standard JTA APIs to demarcate your transaction and you are done!

Example 5. Demarcate your transaction with standalone JTA
//note that you must start the transaction before creating the EntityManager
//or else call entityManager.joinTransaction()
transactionManager.begin();

final EntityManager em = emf.createEntityManager();

Poem poem = new Poem();
poem.setName("L'albatros");
em.persist(poem);

transactionManager.commit();

em.clear();

transactionManager.begin();

poem = em.find(Poem.class, poem.getId());
assertThat(poem).isNotNull();
assertThat(poem.getName()).isEqualTo("L'albatros");
em.remove(poem );

transactionManager.commit();

em.close();

That was not too hard, was it? Note that application frameworks like the Spring Framework should be able to initialize the transaction manager and call it to demarcate transactions for you. Check their respective documentation.

4.2.3. Without JTA

While this approach works today, it does not ensure that operations are done transactionally and hence won’t be able to rollback your work. This will change in the future but in the mean time, such an environment is not recommended.

For NoSQL datastores not supporting transactions, this is less of a concern.

4.3. Configuration options

The most important options when configuring Hibernate OGM are related to the datastore. They are explained in NoSQL datastores.

Otherwise, most options from Hibernate ORM and Hibernate Search are applicable when using Hibernate OGM. You can pass them as you are used to do either in your persistence.xml file, your hibernate.cfg.xml file or programmatically.

More interesting is a list of options that do not apply to Hibernate OGM and that should not be set:

  • hibernate.dialect

  • hibernate.connection.* and in particular hibernate.connection.provider_class

  • hibernate.show_sql and hibernate.format_sql

  • hibernate.default_schema and hibernate.default_catalog

  • hibernate.use_sql_comments

  • hibernate.jdbc.*

  • hibernate.hbm2ddl.auto and hibernate.hbm2ddl.import_file

Hibernate Search integrates with Hibernate OGM just like it does with Hibernate ORM. The Hibernate Search version tested is 5.6.3.Final. Add the dependency to your project - the group id is org.hibernate and artifact id hibernate-search-orm.

Then configure where you want to store your indexes, map your entities with the relevant index annotations and you are good to go. For more information, simply check the Hibernate Search reference documentation.

In Storing a Lucene index in Infinispan we’ll discuss how to store your Lucene indexes in Infinispan. This is useful even if you don’t plan to use Infinispan as your primary data store.

Hibernate OGM requires Hibernate Search on the classpath only when you need to run JPQL or HQL queries with some datastores. This is because some datastores don’t have a query language or we don’t support it yet. In this situation you need to index the entities that you want to query and Hibernate OGM will convert the queries in Lucene queries. Check the paragraph related to the datastore of your choice to see if it requires Hibernate Search or not.

4.5. How to package Hibernate OGM applications for WildFly 10

Provided you’re deploying on WildFly, there is an additional way to add the OGM dependencies to your application.

In WildFly, class loading is based on modules; this system defines explicit, non-transitive dependencies on other modules.

Modules allow to share the same artifacts across multiple applications, making deployments smaller and quicker, and also making it possible to deploy multiple different versions of any library.

More details about modules are described in Class Loading in WildFly.

When deploying a JPA application on WildFly, you should be aware that there are some additional useful configuration properties defined by the WildFly JPA subsystem. These are documented in WildFly JPA Reference Guide.

If you apply the following instructions you can create small and efficient deployments which do not include any dependency, as you can include your favourite version of Hibernate OGM directly to the collection of container provided libraries.

4.5.1. Packaging Hibernate OGM applications for WildFly 10

When using WildFly several of the technologies it includes are automatically enabled. For example Hibernate ORM is made available to your applications if your persistence.xml defines a persistence unit using Hibernate as persistence provider (or is not specifying any provider, as Hibernate is the default one).

Similarly, Hibernate Search is automatically activated and made available on the user’s application classpath if and when the application server detects the need for it. This is the default behaviour, but you are in control and can override this all; see the WildFly JPA Reference Guide for a full list of properties you can explicitly set.

WildFly 10 however does not include Hibernate OGM and the compatible Hibernate ORM and Hibernate Search versions and it will require some configuration to make everything works.

Set up WildFly after downloading the modules

You can download the pre-packaged module ZIP for this version of Hibernate OGM from:

Hibernate OGM 5.2.0.Beta1 requires Hibernate ORM 5.1.10.Final and Hibernate Search 5.6.3.Final.

These versions are not included in WildFly 10 but you can download the pre-packaged ZIP modules from Maven Central:

Once you have the archives, you need to unpack them into the modules folder of your WildFly 10 installation. The modules included are:

  • org.hibernate.ogm, the core Hibernate OGM library.

  • org.hibernate.ogm.<%DATASTORE%>, one module for each datastore, with <%DATASTORE%> being one of infinispan, mongodb etc.

  • org.hibernate.orm, the Hibernate ORM libraries.

  • org.hibernate.search, the Hibernate Search libraries.

  • Several shared dependencies such as org.hibernate.hql:<%VERSION%> (containing the query parser) and others

The module slot to use for Hibernate OGM 5.2.0.Beta1 is 5.2 as the format of the slot name does not include the "micro" part of the project version.

You will also need to set the property wildfly.jpa.hibernate.search.module to org.hibernate.search.orm:5.6.3.Final-orm51 in your persistence.xml. This way your application will use the right Hibernate ORM and Hibernate Search version (and not the one shipped with WildFly).

Example 6. Property for enabling a Hibernate Search and Hibernate ORM version compatible with Hibernate OGM
<property name="wildfly.jpa.hibernate.search.module" value="org.hibernate.search.orm:5.6.3.Final-orm51"/>

The Hibernate Search documentation explains the details of downloading and deploying a custom version: Update and activate latest Hibernate Search version in WildFly.

Set up WildFly using Maven

If you are using Maven, you can download and set up your WildFly with the following snippet:

Example 7. Maven example to prepare a WildFly installation for integration tests
<plugin>
    <artifactId>maven-dependency-plugin</artifactId>
    <executions>
        <execution>
            <id>unpack</id>
            <phase>pre-integration-test</phase>
            <goals>
                <goal>unpack</goal>
            </goals>
            <configuration>
                <artifactItems>
                    <!-- Download and unpack WildFly -->
                    <artifactItem>
                        <groupId>org.wildfly</groupId>
                        <artifactId>wildfly-dist</artifactId>
                        <version>10.1.0.Final</version>
                        <type>zip</type>
                        <overWrite>false</overWrite>
                        <outputDirectory>${project.build.directory}</outputDirectory>
                    </artifactItem>

                    <!-- Download and unpack Hibernate ORM modules -->
                    <artifactItem>
                        <groupId>org.hibernate</groupId>
                        <artifactId>hibernate-orm-modules</artifactId>
                        <version>5.1.10.Final</version>
                        <classifier>wildfly-10-dist</classifier>
                        <type>zip</type>
                        <overWrite>false</overWrite>
                        <outputDirectory>
                            ${project.build.directory}/wildfly-10.1.0.Final/modules
                        </outputDirectory>
                    </artifactItem>

                    <!-- Download and unpack Hibernate Search modules -->
                    <artifactItem>
                        <groupId>org.hibernate</groupId>
                        <artifactId>hibernate-search-modules</artifactId>
                        <version>5.6.3.Final</version>
                        <classifier>wildfly-10-dist</classifier>
                        <type>zip</type>
                        <overWrite>false</overWrite>
                        <outputDirectory>
                            ${project.build.directory}/wildfly-10.1.0.Final/modules
                        </outputDirectory>
                    </artifactItem>

                    <!-- Download and unpack Hibernate OGM modules -->
                    <artifactItem>
                        <groupId>org.hibernate.ogm</groupId>
                        <artifactId>hibernate-ogm-modules</artifactId>
                        <classifier>wildfly-10-dist</classifier>
                        <version>5.2.0.Beta1</version>
                        <type>zip</type>
                        <overWrite>false</overWrite>
                        <outputDirectory>
                            ${project.build.directory}/wildfly-10.1.0.Final/modules
                        </outputDirectory>
                    </artifactItem>
                </artifactItems>
            </configuration>
        </execution>
    </executions>
</plugin>

Now that WildFly is ready, you can include the dependencies in your application in two ways:

Include dependencies using the manifest

Add this entry to the MANIFEST.MF in your archive (replace <%DATASTORE%> with the right value for your chosen datastore):

Dependencies: org.hibernate.ogm:5.2 services, org.hibernate.ogm.<%DATASTORE%>:5.2 services
Include dependencies using jboss-deployment-structure.xml

This is a JBoss-specific descriptor. Add a WEB-INF/jboss-deployment-structure.xml in your archive with the following content (replace <%DATASTORE%> with the right value for your chosen datastore):

<jboss-deployment-structure>
    <deployment>
        <dependencies>
            <module name="org.hibernate.ogm" slot="5.2" services="export" />
            <module name="org.hibernate.ogm.<%DATASTORE%>" slot="5.2" services="export" />
        </dependencies>
    </deployment>
</jboss-deployment-structure>

More information about the descriptor can be found in the WildFly documentation.

4.5.2. Configure your persistence.xml to use your choice of persistence provider

WildFly will by default attempt to guess which Persistence Provider you need by having a look at the provider section of the persistence.xml.

4.5.3. Using the Hibernate OGM modules with Infinispan

The Infinispan project also provides custom modules for WildFly 10. Hibernate OGM modules require these modules if you’re planning to use the Hibernate OGM / Infinispan combination on WildFly.

This release of Hibernate OGM was tested exclusively with Infinispan version 9.1.1.Final; the Infinispan project generally attempts to maintain the same API and integration points within the same major.minor version, so a micro version update should be safe but is untested.

In case you want to experiment with a more significant version upgrade, you will need to edit the modules of Hibernate OGM: the module identifiers are hardcoded in the XML files representing the module.

Download the Infinispan modules pack for WildFly 10 from here:

Then similarly to what you did with the Hibernate OGM modules zip, unpack this one too in your modules directory within the application server.

5. Map your entities

This section mainly describes the specificities of Hibernate OGM mappings. It’s not meant to be a comprehensive guide to entity mappings, the complete guide is Hibernate ORM’s documentation: after all Hibernate OGM is Hibernate ORM.

5.1. Supported entity mapping

Pretty much all entity related constructs should work out of the box in Hibernate OGM. @Entity, @Table, @Column, @Enumerated, @Temporal, @Cacheable and the like will work as expected. If you want an example, check out Getting started with Hibernate OGM or the documentation of Hibernate ORM. Let’s concentrate of the features that differ or are simply not supported by Hibernate OGM.

Hibernate OGM supports the following inheritance strategies: * InheritanceType.TABLE_PER_CLASS * InheritanceType.SINGLE_TABLE

If you feel the need to support other strategies, let us know (see How to contribute).

JPA annotations refer to tables but the kind of abstraction the database will use depends on the nature of the NoSQL datastore you are dealing with. For example, in MongoDB a table is mapped as a document.

You can find more details about the way entities are stored in the corresponding mapping section of the datastore you are using.

Secondary tables are not supported by Hibernate OGM at the moment. If you have needs for this feature, let us know (see How to contribute).

Queries are partially supported, you will find more information in the query chapter.

All standard JPA id generators are supported: IDENTITY, SEQUENCE, TABLE and AUTO. If you need support for additional generators, let us know (see How to contribute).

Some NoSQL databases can not provide an efficient implementation for IDENTITY or SEQUENCE, for these cases we recommend you use a UUID based generator. For example on Infinispan (in embedded mode) using IDENTITY is possible but it will require using cluster wide coordination to maintain the counters, which is not going to perform very well.

@Entity
public class Breed {

    @Id @GeneratedValue(generator = "uuid")
    @GenericGenerator(name="uuid", strategy="uuid2")
    public String getId() { return id; }
    public void setId(String id) { this.id = id; }
    private String id;

    public String getName() { return name; }
    public void setName(String name) { this.name = name; }
    private String name;
}

5.2. Supported Types

Most Java built-in types are supported at this stage. However, custom types (@Type) are not supported.

Here is a list of supported Java types:

  • Boolean

  • Byte

  • Byte Array

  • Calendar

  • Class

  • Date

  • Double

  • Integer

  • Long

  • Short

  • Float

  • Character

  • String

  • BigDecimal (mapped as scientific notation)

  • BigInteger

  • Url (as described by RFC 1738 and returned by toString of the Java URL type)

  • UUID stored as described by RFC 4122

  • Enums

Let us know if you need more type support How to contribute

5.3. Supported association mapping

All association types are supported (@OneToOne, @OneToMany, @ManyToOne, @ManyToMany). Likewise, all collection types are supported (Set, Map, List). The way Hibernate OGM stores association information is however quite different than the traditional RDBMS representation. Each chapter dedicated to a datastore describes how associations are persisted, make sure to check them out.

Not all types of associations can be mapped efficiently on all datastores: this will depend on the specific capabilities of the NoSQL technology being used. For example the key/value stores will have all of the association navigation for a given entity stored in a single key. If your collection is made of 1 million elements, Hibernate OGM will have to store 1 million tuples in the association key. For example the Infinispan Embedded dialect suffers from this limitation as it’s treated as a pure key/value store, while the Infinispan Remote dialect does not as it is more similar to a document store.

6. Hibernate OGM APIs

Hibernate OGM has very few specific APIs. For the most part, you will interact with it via either:

  • the JPA APIs

  • the native Hibernate ORM APIs

This chapter will only discuss the Hibernate OGM specific behaviors regarding these APIs. If you need to learn JPA or the native Hibernate APIs, check out the Hibernate ORM documentation.

6.1. Bootstrap Hibernate OGM

We already discussed this subject earlier, have a look at Configure and start Hibernate OGM for all the details.

As a reminder, it basically boils down to either:

  • set the right persistence provider in your persistence.xml file and create an EntityManagerFactory the usual way

  • start via the Hibernate ORM native APIs using StandardServiceRegistryBuilder and MetadataSources to boot a SessionFactory

6.2. JPA and native Hibernate ORM APIs

You know of the Java Persistence and Hibernate ORM native APIs? You are pretty much good to go. If you need a refresher, make sure you read the Hibernate ORM documentation.

A few things are a bit different though, let’s discuss them.

Most of the EntityManager and Session contracts are supported. Here are the few exceptions:

  • Session.createCriteria: criteria queries are not yet supported in Hibernate OGM

  • Session.createFilter: queries on collections are not supported yet

  • Session 's enableFilter, disableFilter etc: query filters are not supported at the moment

  • doWork and doReturningWork are not implemented as they rely on JDBC connections - see OGM-694

  • Session 's stored procedure APIs are not supported

  • Session 's natural id APIs are not yet supported

  • Session.lock is not fully supported at this time

  • EntityManager 's criteria query APIs are not supported

  • EntityManager 's stored procedure APIs are not supported - see OGM-695

  • EntityManager.lock is not fully supported at this time

  • see Query your entities to know what is supported for JPQL and native queries

6.2.1. Accessing the OgmSession API

To execute NoSQL native queries, one approach is to use OgmSession#createNativeQuery. You can read more about it in Using the native query language of your NoSQL. But let’s see how to access an OgmSession instance.

From JPA, use the unwrap method of EntityManager

Example 8. Get to an OgmSession from an EntityManager
EntityManager entityManager = ...
OgmSession ogmSession = entityManager.unwrap(OgmSession.class);
NoSQLQuery query = ogmSession.createNativeQuery(...);

In the Hibernate native API case, you should already have access to an OgmSession. The OgmConfiguration you used returns an OgmSessionFactory. This factory in turns produces OgmSession.

Example 9. Get to an OgmSession with Hibernate ORM native APIs
StandardServiceRegistry registry = new StandardServiceRegistryBuilder()
    .applySetting( OgmProperties.ENABLED, true )
    .build();

OgmSessionFactory ogmSessionFactory = new MetadataSources( registry )
    .buildMetadata()
    .getSessionFactoryBuilder()
    .unwrap( OgmSessionFactoryBuilder.class )
    .build();
OgmSession ogmSession = ogmSessionFactory.openSession();
NoSQLQuery query = ogmSession.createNativeQuery(...);

6.3. On flush and transactions

Even though some underlying NoSQL datastores do not support transaction, it is important to demarcate transaction via the Hibernate OGM APIs. Let’s see why.

Hibernate does pile up changes for as long as it can before pushing them down to the datastore. This opens up the doors to huge optimizations (avoiding duplication, batching operations etc). You can force changes to be sent to the datastore by calling Session.flush or EntityManager.flush. In some situations - for example before some queries are executed -, Hibernate will flush automatically. It will also flush when the transaction demarcation happens (whether there is a real transaction or not).

The best approach is to always demarcate the transaction as shown below. This avoids the needs to manually call flush and will offer future opportunities for Hibernate OGM.

Example 10. Explicitly demarcating transactions

Here is how you do outside of a JTA environment.

Session session = ...

Transaction transaction = session.beginTransaction();
try {
    // do your work
    transaction.commit(); // will flush changes to the datastore
catch (Exception e) {
    transaction.rollback();
}

// or in JPA
EntityManager entityManager = ...
EntityTransaction transaction = entityManager.getTransaction();
try {
    // do your work
    transaction.commit(); // will flush changes to the datastore
}
catch (Exception e) {
    transaction.rollback();
}

Inside a JTA environment, either the container demarcates the transaction for you and Hibernate OGM will transparently join that transaction and flush at commit time. Or you need to manually demarcate the transaction. In the latter case, it is best to start / stop the transaction before retrieving the Session or EntityManager as shown below. The alternative is to call the EntityManager.joinTransaction() once the transaction has started.

transactionManager.begin();
Session session = sessionFactory.openSession();
// do your work
transactionManager.commit(); // will flush changes to the datastore

// or in JPA
transactionManager.begin();
EntityManager entityManager = entityManagerFactory.createEntityManager();
// do your work
transactionManager.commit(); // will flush changes to the datastore

6.3.1. Acting upon errors during application of changes

The error compensation API described in the following section is an experimental feature. It will be enriched with additional features over time. This might require changes to existing method signatures and thus may break code using a previous version of the API.

Please let us know about your usage of the API and your wishes regarding futher capabilities!

If an error occurs during flushing a set of changes, some data changes may already have been applied in the datastore. If the store is non-transactional, there is no way to rollback (undo) these changes if they were already flushed. In this case it is desirable to know which changes have been applied and which ones failed in order to take appropriate action.

Hibernate OGM provides an error compensation API for this purpose. By implementing the org.hibernate.ogm.failure.ErrorHandler interface, you will be notified if

  • an interaction between the Hibernate OGM engine and the grid dialect failed

  • a rollback of the current transaction was triggered

Use cases for the error compensation API include:

  • Logging all applied operations

  • Retrying a failed operation e.g. after timeouts

  • Making an attempt to compensate (apply an inverse operation) applied changes

In its current form the API lays the ground for manually performing these and similar tasks, but we envision a more automated approach in future versions, e.g. for automatic retries of failed operations or the automatic application of compensating operations.

Let’s take a look at an example:

Example 11. Custom ErrorHandler implementation
public class ExampleErrorHandler extends BaseErrorHandler {

    @Override
    public void onRollback(RollbackContext context) {
        // write all applied operations to a log file
        for ( GridDialectOperation appliedOperation : context.getAppliedGridDialectOperations() ) {
            switch ( appliedOperation.getType() ) {
                case INSERT_TUPLE:
                    EntityKeyMetadata entityKeyMetadata = appliedOperation.as( InsertTuple.class ).getEntityKeyMetadata();
                    Tuple tuple = appliedOperation.as( InsertTuple.class ).getTuple();

                    // write EKM and tuple to log file...
                    break;
                case REMOVE_TUPLE:
                    // ...
                    break;
                case ...
                    // ...
                    break;
            }
        }
    }

    @Override
    public ErrorHandlingStrategy onFailedGridDialectOperation(FailedGridDialectOperationContext context) {
        // Ignore this exception and continue
        if ( context.getException() instanceof TupleAlreadyExistsException ) {
            GridDialectOperation failedOperation = context.getFailedOperation();
            // write to log ...

            return ErrorHandlingStrategy.CONTINUE;
        }
        // But abort on all others
        else {
            return ErrorHandlingStrategy.ABORT;
        }
    }
}

The onRollback() method - which is called when the transaction is rolled back (either by the user or by the container) - shows how to iterate over all methods applied prior to the rollback, examine their specific type and e.g. write them to a log file.

The onFailedGridDialectOperation() method is called for each specific datastore operation failing. It lets you decide whether to continue ignoring the failure, retry or abort the operation. If ABORT is returned, the causing exception will be re-thrown, eventually causing the current transaction to be rolled back. If CONTINUE is returned, that exception will be ignored, causing the current transaction to continue.

The decision whether to abort or continue can be based on the specific exception type or on the grid dialect operation which caused the failure. In the example all exceptions of type TupleAlreadyExistsException are ignored, whereas all other exceptions cause the current flush cycle to be aborted. You also could react to datastore-specific exceptions such as MongoDB’s MongoTimeoutException, if needed.

Note that by extending the provided base class BaseErrorHandler rather than implementing the interface directly, you only need to implement those callback methods you are actually interested in. The implementation will also not break if further callback methods are added to the ErrorHandler interface in future releases.

Having implemented the error handler, it needs to be registered with Hibernate OGM. To do so, specify it using the property hibernate.ogm.error_handler, e.g. as a persistence unit property in META-INF/persistence.xml:

<property name="hibernate.ogm.error_handler" value="com.example.ExampleErrorHandler"/>

6.4. SPIs

Some of the Hibernate OGM public contracts are geared towards either integrators or implementors of datastore providers. They should not be used by a regular application. These contracts are named SPIs and are in a .spi package.

To keep improving Hibernate OGM, we might break these SPIs between versions. If you plan on writing a datastore, come and talk to us.

Non public contracts are stored within a .impl package. If you see yourself using one of these classes, beware that we can break these without notice.

7. Query your entities

Once your data is in the datastore, it’s time for some query fun! With Hibernate OGM, you have a few alternatives that should get you covered:

  • Use JPQL - only for simple queries for now

  • Use the NoSQL native query mapping the result as managed entities

  • Use Hibernate Search queries - primarily full-text queries

7.1. Using JPQL

For Hibernate OGM, we developed a brand new JPQL parser which is already able to convert simple queries into the native underlying datastore query language (e.g. MongoQL for MongoDB, CypherQL for Neo4J, etc). This parser can also generate Hibernate Search queries for datastores that do not support a query language.

For datastores like Infinispan that require Hibernate Search to execute JPQL queries, the following preconditions must be met:

  • no join, aggregation, or other relational operations are implied

  • the entity involved in the query must be indexed

  • the properties involved in the predicates must be indexed

Here is an example:

@Entity @Indexed
public class Hypothesis {

    @Id
    public String getId() { return id; }
    public void setId(String id) { this.id = id; }
    private String id;

    @Field(analyze=Analyze.NO)
    public String getDescription() { return description; }
    public void setDescription(String description) { this.description = description; }
    private String description;
}

Query query = session
    .createQuery("from Hypothesis h where h.description = :desc")
    .setString("desc", "tomorrow it's going to rain");

Note that the description field is marked as not analysed. This is necessary to support field equality and comparison as defined by JPQL.

You can make use of the following JPQL constructs:

  • simple comparisons using "<", "<=", "=", ">=" and ">"

  • IS NULL and IS NOT NULL

  • the boolean operators AND, OR, NOT

  • LIKE, IN and BETWEEN

  • ORDER BY

In particular and of notice, what is not supported is:

  • cross entity joins

  • JPQL functions in particular aggregation functions like count

  • JPQL update and delete queries

That may sound rather limiting for your use cases so bear with us. This is a hot area we want to improve, please tell us what feature you miss by opening a JIRA or via email. Also read the next section, you will see other alternatives to implement your queries.

Let’s look at some of the queries you can express in JPQL:

Example 12. Some JPQL queries
// query returning an entity based on a simple predicate
select h from Hypothesis h where id = 16

// projection of the entity property
select id, description from Hypothesis h where id = 16

// projection of the embedded properties
select h.author.address.street from Hypothesis h where h.id = 16

// predicate comparing a property value and a literal
from Hypothesis h where h.position = '2'

// negation
from Hypothesis h where not h.id = '13'
from Hypothesis h where h.position <> 4

// conjunction
from Hypothesis h where h.position = 2 and not h.id = '13'

// named parameters
from Hypothesis h where h.description = :myParam

// range query
from Hypothesis h where h.description BETWEEN :start and :end"

// comparisons
from Hypothesis h where h.position < 3

// in
from Hypothesis h where h.position IN (2, 3, 4)

// like
from Hypothesis h where h.description LIKE '%dimensions%'

// comparison with null
from Hypothesis h where h.description IS null

// order by
from Hypothesis h where h.description IS NOT null ORDER BY id
from Helicopter h order by h.make desc, h.name

There are also features that are partially supported:

  • Inner JOIN on an embedded association: works as expected with Neo4j and MongoDB; doesn’t work if your datastore provider implements JPQL queries using Hibernate Search.

  • Projections or filters on properties of an embedded identifier: works as expected with Neo4j and MongoDB; doesn’t work if your datastore provider implements JPQL queries using Hibernate Search.

These are better illustrated by the following example:

Example 13. Entity with embedded collection and supported JPQL queries
@Indexed
@Entity
public class StoryGame {

    @DocumentId
    @EmbeddedId
    @FieldBridge(impl = NewsIdFieldBridge.class)
    private StoryID storyId;

    @ElementCollection
    @IndexedEmbedded
    private List<OptionalStoryBranch> optionalEndings;

    ...

}

@Embeddable
public class StoryID implements Serializable {

    private String title;
    private String author;

    ...
}

@Embeddable
public class OptionalStoryBranch {

    // Analyze.NO for filtering in query
    // Store.YES for projection in query
    @Field(store = Store.YES, analyze = Analyze.NO)
    private String text;

    ...

}

Filter the results using the supported operators will work for all the datastores:

String query =
    "SELECT sg" +
    "FROM StoryGame sg JOIN sg.optionalEndings ending WHERE ending.text = 'Happy ending'"
List<StoryGame> stories = session.createQuery( query ).list();

Projection of properties of an embedded association works with Neo4j and MongoDB, but the other datastores will only return one element from the association. This is due to the fact that Hibernate Search is currently not supporting projection of associations. Here’s an example of a query affected by this:

String query =
     "SELECT ending.text " +
     "FROM StoryGame sg JOIN sg.optionalEndings ending WHERE ending.text LIKE 'Happy%'";
List<String> endings = session.createQuery( query ).list();

Projecting and filtering on embedded id properties works with Neo4j and MongoDB but throws an exception with the other datastores:

String query =
     "SELECT sg.storyId.title FROM StoryGame sg WHERE sg.storyId.title = 'Best Story Ever'";
List<String> title = session.createQuery( query ).list();

It will cause the following exception if the datastore uses Hibernate Search to execute JPQL queries:

org.hibernate.hql.ParsingException: HQL100002: The type [storyId] has no indexed property named title.

In order to reflect changes performed in the current session, all entities affected by a given query are flushed to the datastore prior to query execution (that’s the case for Hibernate ORM as well as Hibernate OGM).

For not fully transactional stores, this can cause changes to be written as a side-effect of running queries which cannot be reverted by a possible later rollback.

Depending on your specific use cases and requirements you may prefer to disable auto-flushing, e.g. by invoking query.setFlushMode(FlushMode.MANUAL). Bear in mind though that query results will then not reflect changes applied within the current session.

7.2. Using the native query language of your NoSQL

Often you want the raw power of the underlying NoSQL query engine. Even if that costs you portability.

Hibernate OGM addresses that problem by letting you express native queries (e.g. in MongoQL or CypherQL) and map the result of these queries as mapped entities.

In JPA, use EntityManager.createNativeQuery. The first form accepts a result class if your result set maps the mapping definition of the entity. The second form accepts the name of a resultSetMapping and lets you customize how properties are mapped to columns by the query. You can also used a predefined named query which defines its result set mapping.

Let’s take a look at how it is done for Neo4J:

Example 14. Various ways to create a native query in JPA
@Entity
@NamedNativeQuery(
   name = "AthanasiaPoem",
   query = "{ $and: [ { name : 'Athanasia' }, { author : 'Oscar Wilde' } ] }",
   resultClass = Poem.class )
public class Poem {

    @Id
    private Long id;

    private String name;

    private String author;

   // getters, setters ...

}

...

javax.persistence.EntityManager em = ...

// a single result query
String query1 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) RETURN n";
Poem poem = (Poem) em.createNativeQuery( query1, Poem.class ).getSingleResult();

// query with order by
String query2 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) " +
                "RETURN n ORDER BY n.name";
List<Poem> poems = em.createNativeQuery( query2, Poem.class ).getResultList();

// query with projections
String query3 = MATCH ( n:Poem ) RETURN n.name, n.author ORDER BY n.name";
List<Object[]> poemNames = (List<Object[]>)em.createNativeQuery( query3 )
                               .getResultList();

// named query
Poem poem = (Poem) em.createNamedQuery( "AthanasiaPoem" ).getSingleResult();

In the native Hibernate API, use OgmSession.createNativeQuery or Session.getNamedQuery. The former form lets you define the result set mapping programmatically. The latter is receiving the name of a predefined query already describing its result set mapping.

Example 15. Hibernate API defining a result set mapping
OgmSession session = ...
String query1 = "{ $and: [ { name : 'Portia' }, { author : 'Oscar Wilde' } ] }";
Poem poem = session.createNativeQuery( query1 )
                      .addEntity( "Poem", Poem.class )
                      .uniqueResult();

Check out each individual datastore chapter for more info on the specifics of the native query language mapping. In particular Neo4J and MongoDB.

Hibernate Search offers a way to index Java objects into Lucene indexes and to execute full-text queries on them. The indexes do live outside your datastore. This offers a few interesting properties in terms of feature set and scalability.

Apache Lucene is a full-text indexing and query engine with excellent query performance. Feature wise, full-text means you can do much more than a simple equality match.

Hibernate Search natively integrates with Hibernate ORM. And Hibernate OGM of course!

Example 16. Adding Hibernate Search artifact to your project via Maven
<dependency>
    <groupId>org.hibernate</groupId>
    <artifactId>hibernate-search-orm</artifactId>
</dependency>
Example 17. Using Hibernate Search for full-text matching
@Entity @Indexed
public class Hypothesis {

    @Id
    public String getId() { return id; }
    public void setId(String id) { this.id = id; }
    private String id;

    @Field(analyze=Analyze.YES)
    public String getDescription() { return description; }
    public void setDescription(String description) { this.description = description; }
    private String description;
}
EntityManager entityManager = ...
//Add full-text superpowers to any EntityManager:
FullTextEntityManager ftem = Search.getFullTextEntityManager(entityManager);

//Optionally use the QueryBuilder to simplify Query definition:
QueryBuilder b = ftem.getSearchFactory()
   .buildQueryBuilder()
   .forEntity(Hypothesis.class)
   .get();

//Create a Lucene Query:
Query lq = b.keyword().onField("description").matching("tomorrow").createQuery();

//Transform the Lucene Query in a JPA Query:
FullTextQuery ftQuery = ftem.createFullTextQuery(lq, Hypothesis.class);

//List all matching Hypothesis:
List<Hypothesis> resultList = ftQuery.getResultList();

Assuming our database contains an Hypothesis instance having description "Sometimes tomorrow we release", that instance will be returned by our full-text query.

Text similarity can be very powerful as it can be configured for specific languages or domain specific terminology; it can deal with typos and synonyms, and above all it can return results by relevance.

Worth noting the Lucene index is a vectorial space of term occurrence statistics: so extracting tags from text, frequencies of strings and correlate this data makes it very easy to build efficient data analysis applications.

While the potential of Lucene queries is very high, it’s not suited for all use cases Let’s see some of the limitations of Lucene Queries as our main query engine:

  • Lucene doesn’t support Joins. Any to-One relations can be mapped fine, and the Lucene community is making progress on other forms, but restrictions on OneToMany or ManyToMany can’t be implemented today.

  • Since we apply changes to the index at commit time, your updates won’t affect queries until you commit (we might improve on this).

  • While queries are extremely fast, write operations are not as fast (but we can make it scale).

For a complete understanding of what Hibernate Search can do for you and how to use it, go check the Hibernate Search reference documentation.

7.4. Using the Criteria API

At this time, we have not implemented support for the Criteria APIs (neither Hibernate native nor JPA).

8. NoSQL datastores

Currently Hibernate OGM supports the following NoSQL datastores:

  • Map: stores data in an in-memory Java map to store data. Use it only for unit tests.

  • Infinispan Embedded: stores data into Infinispan (data grid) participating directly in the cluster

  • Infinispan Remote: (also called Hot Rod, the name of the Infinispan client) stores data into Infinispan by connecting as an Hot Rod client

    • at this stage, this datastore is experimental

  • Ehcache: stores data into Ehcache (cache)

  • MongoDB: stores data into MongoDB (document store)

  • Neo4j: stores data into Neo4j (graph database)

  • CouchDB: stores data into CouchDB (document store)

    • at this stage, this datastore is experimental

  • Redis: stores data into Redis (key-value store)

    • at this stage, this datastore is experimental

More are planned, if you are interested, come talk to us (see How to get help and contribute on Hibernate OGM).

For each datastore, Hibernate OGM has specific integration code called a datastore provider. All are in a dedicated maven module, you simply need to depend on the one you use.

Hibernate OGM interacts with NoSQL datastores via two contracts:

  • a DatastoreProvider which is responsible for starting and stopping the connection(s) with the datastore and prop up the datastore if needed

  • a GridDialect which is responsible for converting an Hibernate OGM operation into a datastore specific operation

8.1. Using a specific NoSQL datastore

Only a few steps are necessary:

  • add the datastore provider module to your classpath

  • ask Hibernate OGM to use that datastore provider

  • configure the URL or configuration to reach the datastore

Adding the relevant Hibernate OGM module in your classpath looks like this in Maven:

<dependency>
    <groupId>org.hibernate.ogm</groupId>
    <artifactId>hibernate-ogm-infinispan</artifactId>
    <version>5.2.0.Beta1</version>
</dependency>

The module names are hibernate-ogm-infinispan, hibernate-ogm-infinispan-remote, hibernate-ogm-ehcache, hibernate-ogm-mongodb, hibernate-ogm-neo4j, hibernate-ogm-couchdb, hibernate-ogm-redis and hibernate-ogm-cassandra. The map datastore is included in the Hibernate OGM engine module.

Next, configure which datastore provider you want to use. This is done via the hibernate.ogm.datastore.provider option. Possible values are:

  • specific shortcuts (preferable): map (only to be used for unit tests), infinispan_embedded, infinispan_remote, ehcache, mongodb, neo4j_embedded, neo4j_http, neo4j_bolt, couchdb_experimental or redis_experimental

  • the fully qualified class name of a DatastoreProvider implementation

When bootstrapping a session factory or entity manager factory programmatically, you should use the constants declared on org.hibernate.ogm.cfg.OgmProperties to specify configuration properties such as hibernate.ogm.datastore.provider.

In this case you also can specify the provider in form of a class object of a datastore provider type or pass an instance of a datastore provider type:

Map<String, Object> properties = new HashMap<String, Object>();

// pass the type
properties.put( OgmProperties.DATASTORE_PROVIDER, MongoDBDatastoreProvider.class );

EntityManagerFactory emf = Persistence.createEntityManagerFactory( "my-pu", properties );

By default, a datastore provider chooses the grid dialect transparently but you can override this setting with the hibernate.ogm.datastore.grid_dialect option. Use the fully qualified class name of the GridDialect implementation. Most users should ignore this setting entirely.

Let’s now look at the specifics of each datastore provider. How to configure it further, what mapping structure is used and more.

9. Infinispan

Infinispan is an open source in-memory data grid focusing on high performance. As a data grid, you can deploy it on multiple servers - referred to as nodes - and connect to it as if it were a single storage engine: it will cleverly distribute both the computation effort and the data storage.

It is trivial to setup on a single node and Hibernate OGM knows how to boot one, so you can easily try it out. But Infinispan really shines in multiple node deployments: you will need to configure some networking details but nothing changes in terms of application behaviour, while performance and data size can scale linearly.

From all its features we will only describe those relevant to Hibernate OGM; for a complete description of all its capabilities and configuration options, refer to the Infinispan project documentation at infinispan.org.

9.1. Why use Hibernate OGM with Infinispan?

Infinispan provides great scalability and elasticity features but it can have a steep learning curve.

If you are already familiar with the JPA API you will be able to store your data in Infinispan quickly and you will also benefit from the optimizations that a framework like Hibernate OGM can apply under the hood.

In particular:

  • you can get started without having to learn Protobuf first

  • no need to learn the Infinispan API

  • Hibernate OGM will setup and manage the Hot Rod client for you

  • same API as Hibernate ORM, meaning that you can use the same tools

You will still need to learn about Infinispan, all its capabilities and how to configure them to reach your application top performance, but you can get a proof of concept done quickly with the example configuration.

9.2. Infinispan: Choosing between Embedded Mode and Hot Rod

Java applications can use Infinispan in two fundamentally different ways:

  • Run Infinispan in Embedded Mode.

  • Connect to an Infinispan Server using an Hot Rod client.

Hibernate OGM supports connecting in either mode, but since the APIs and capabilities are different in the two modes, it provides two different modules each having its own set of configuration options and features.

Running Infinispan in Embedded Mode implies that the Infinispan node is running in the same JVM as the code using it. The benefit is that some data (or all data) will be stored on the same JVM, making reads of this data extremely fast and efficient as there won’t be RPCs to other systems. Write operations will still need to issue some coordination RPCs but they also benefit from a reduction of necessary operations.

However the very fact that some data is stored in the same JVM is also the drawback of this choice: this typically implies having to configure your JVM for larger heap sizes, which are harder to tune for optimal performance. Other system parameters might also need to be configured as this JVM node is now to be treated as a "data holding node" rather than a stateless app node. Some architects and system administrators will not like that.

When connecting to an Infinispan Server over the Hot Rod client, the architecture is similar to having Hibernate connect to traditional database: the data is stored on the Infinispan Server nodes, and Hibernate OGM uses a client with a pool of TCP connections to talk to the server.

Another important difference, is that when connecting to Infinispan Server via Hot Rod the data is encoded using Google Protobuf, which requires a schema. This schema is auto-generated by Hibernate OGM.

Having a Protobuf Schema makes it possible to evolve the schema in non-destructive ways, and makes it possible for other clients to access the data - even clients written in other programming languages.

Most introductory tutorials of Hibernate OGM focus on Infinispan in Embedded Mode because in this mode OGM can start its own embedded Infinispan node, using a simple, local only Infinispan configuration.

When using Infinispan Server instead, you’ll need to download the server distribution, unpack and start it, then set the Hibernate OGM configuration properties so that the integrated Hot Rod client knows how to connect to it.

Advanced performance options & interaction with Hibernate 2nd level caching

When using Infinispan in Embedded Mode, and it’s caches are configured in REPLICATION Mode, all nodes will contain a full replica of the database: write performance won’t scale but your reads will be very fast and scale up linearly with the size of the cluster, making usage of Hibernate’s 2nd level cache redundant.

When configuring Infinispan in DISTRIBUTED cache mode, each of your nodes will have a local copy of a slice of your data; remember you can tune how large the section should be with various Infinispan configuration options (such as numOwners), and you could combine this with Hibernate’s 2nd level caching and/or enable Infinispan’s 1st level caching.

You can even combine Infinispan with having it passivate to other storage systems such as a RDBMs or another NoSQL engine; such storage can be configured to be asynchronous. This option is available to both Infinispan Embedded and Infinispan Server; it’s even possible to use a light layer of Infinispan Embedded - containing a small data set - and have it backed by an Infinispan Server cluster to expand its storage capabilities without having to enlarge heap size too much on the embedded, application nodes.

Finally, remember that options such as replication vs distribution (CacheMode) and passivation to additional storage (CacheStores) can be configured differently for each of Infinispan caches.

9.3. Hibernate OGM & Infinispan Embedded

Let’s see how to configure and use Hibernate OGM with Infinispan in Embedded Mode.

For usage of Infinispan Server over Hot Rod, skip to Hibernate OGM & Infinispan Server over Hot Rod.

9.3.1. Configure Hibernate OGM for Infinispan Embedded

You configure Hibernate OGM and Infinispan in two steps basically:

  • Add the dependencies to your classpath

  • And then choose one of:

    • Use the default Infinispan configuration (no action needed)

    • Point to your own configuration resource file

    • Point to a JNDI name of an existing instance of an Infinispan CacheManager

  • If you need to run JPQL or HQL queries, add Hibernate Search on the classpath (Using Hibernate Search)

Note that, except when using JNDI, Hibernate OGM will bootstrap and Infinispan Embedded node in your same JVM, and terminate it on shutdown of the Hibernate instance.

If you have Hibernate OGM boot Infinispan using a clustered configuration, this might automatically join the cluster of other Infinispan nodes running in your network: the default is automatic discovery!

9.3.2. Adding Infinispan dependencies

To add the dependencies for the Hibernate OGM extensions for Infinispan Embedded via Maven, add the following module:

<dependency>
    <groupId>org.hibernate.ogm</groupId>
    <artifactId>hibernate-ogm-infinispan</artifactId>
    <version>5.2.0.Beta1</version>
</dependency>

If you’re not using a dependency management tool, copy all the dependencies from the distribution in the directories:

  • /lib/required

  • /lib/infinispan

  • Optionally - depending on your container - you might need some of the jars from /lib/provided

9.3.3. Infinispan specific configuration properties

The advanced configuration details of an Infinispan Cache are defined in an Infinispan specific XML configuration file; the Hibernate OGM properties are simple and usually just point to this external resource.

To use the default configuration provided by Hibernate OGM - which is a good starting point for new users - you don’t have to set any property.

Hibernate OGM properties for Infinispan
hibernate.ogm.datastore.provider

Set it to infinispan_embedded to use Infinispan as the datastore provider in embedded mode.

hibernate.ogm.infinispan.cachemanager_jndi_name

If you have an Infinispan EmbeddedCacheManager registered in JNDI, provide the JNDI name and Hibernate OGM will use this instance instead of starting a new CacheManager. This will ignore any further configuration properties as Infinispan is assumed being already configured. Infinispan can typically be pushed to JNDI via WildFly, Spring or Seam.

hibernate.ogm.infinispan.configuration_resource_name

Should point to the resource name of an Infinispan configuration file. This is ignored in case JNDI lookup is set. Defaults to org/hibernate/ogm/datastore/infinispan/default-config.xml.

hibernate.ogm.datastore.keyvalue.cache_storage

The strategy for persisting data in Infinispan. The following two strategies exist (values of the org.hibernate.ogm.datastore.keyvalue.options.CacheMappingType enum):

  • CACHE_PER_TABLE: A dedicated cache will be used for each entity type, association type and id source table.

  • CACHE_PER_KIND: Three caches will be used: one cache for all entities, one cache for all associations and one cache for all id sources.

Defaults to CACHE_PER_TABLE. It is the recommended strategy as it makes it easier to target a specific cache for a given entity.

hibernate.ogm.datastore.create_database

If set to true Hibernate OGM will create any missing Cache definitions on the Infinispan Server. This requires the Infinispan Server configuration to have a default configuration defined, as this will be copied to the newly defined caches. If set to false an exception is thrown when a Cache is expected but not explicitly configured on the server. Defaults to false.

When bootstrapping a session factory or entity manager factory programmatically, you should use the constants accessible via org.hibernate.ogm.datastore.infinispan.InfinispanProperties when specifying the configuration properties listed above.

Common properties shared between stores are declared on OgmProperties (a super interface of InfinispanProperties).

For maximum portability between stores, use the most generic interface possible.

9.3.4. Cache names used by Hibernate OGM

Depending on the cache mapping approach, Hibernate OGM will either:

  • store each entity type, association type and id source table in a dedicated cache very much like what Hibernate ORM would do. This is the CACHE_PER_TABLE approach.

  • store data in three different caches when using the CACHE_PER_KIND approach:

    • ENTITIES: is going to be used to store the main attributes of all your entities.

    • ASSOCIATIONS: stores the association information representing the links between entities.

    • IDENTIFIER_STORE: contains internal metadata that Hibernate OGM needs to provide sequences and auto-incremental numbers for primary key generation.

The preferred strategy is CACHE_PER_TABLE as it offers both more fine grained configuration options and the ability to work on specific entities in a more simple fashion.

In the following paragraphs, we will explain which aspects of Infinispan you’re likely to want to reconfigure from their defaults. All attributes and elements from Infinispan which we don’t mention are safe to ignore. Refer to the Infinispan User Guide for the guru level performance tuning and customizations.

An Infinispan configuration file is an XML file complying with the Infinispan schema; the basic structure is shown in the following example:

Example 18. Simple example of an Infinispan configuration file
<?xml version="1.0" encoding="UTF-8"?>
<infinispan
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="urn:infinispan:config:9.1 http://www.infinispan.org/schemas/infinispan-config-9.1.xsd"
    xmlns="urn:infinispan:config:9.1">

    <cache-container name="HibernateOGM" default-cache="DEFAULT">

        <!-- Default cache settings -->
        <local-cache name="DEFAULT">
            <transaction mode="NON_DURABLE_XA" />
        </local-cache>

        <local-cache name="User"/>

        <local-cache name="Order"/>

        <local-cache name="associations_User_Order"/>

    </cache-container>
</infinispan>

There are global settings that can be set before the cache_container section. These settings will affect the whole instance; mainly of interest for Hibernate OGM users is the jgroups element in which we will set JGroups configuration overrides.

Inside the cache-container section are defined explicit named caches and their configurations as well as the default cache (named DEFAULT here) if we want to affect all named caches. This is where we will likely want to configure clustering modes, eviction policies and CacheStores.

9.3.5. Manage data size

In its default configuration Infinispan stores all data in the heap of the JVM; in this barebone mode it is conceptually not very different than using a HashMap: the size of the data should fit in the heap of your VM, and stopping/killing/crashing your application will get all data lost with no way to recover it.

To store data permanently (out of the JVM memory) a CacheStore should be enabled. The Infinispan project provides many CacheStore implementations; a simple one is the "Single File Store" which is able to store data in simple binary files, on any read/write mounted filesystem; You can find many more implementations to store your data in anything from JDBC connected relational databases, other NoSQL engines such as MongoDB and Cassandra, or even delegate to other Infinispan clusters. Finally, implementing a custom CacheStore is quite easy.

To limit the memory consumption of the precious heap space, you can activate a passivation or an eviction policy; again there are several strategies to play with, for now let’s just consider you’ll likely need one to avoid running out of memory when storing too many entries in the bounded JVM memory space; of course you don’t need to choose one while experimenting with limited data sizes: enabling such a strategy doesn’t have any other impact in the functionality of your Hibernate OGM application (other than performance: entries stored in the Infinispan in-memory space is accessed much quicker than from any CacheStore).

A CacheStore can be configured as write-through, committing all changes to the CacheStore before returning (and in the same transaction) or as write-behind. A write-behind configuration is normally not encouraged in storage engines, as a failure of the node implies some data might be lost without receiving any notification about it, but this problem is mitigated in Infinispan because of its capability to combine CacheStore write-behind with a synchronous replication to other Infinispan nodes.

Example 19. Enabling a FileCacheStore and eviction
<local-cache name="User">
    <transaction mode="NON_DURABLE_XA" />
    <eviction strategy="LIRS" max-entries="2000"/>
    <persistence passivation="true">
        <file-store
           shared="false"
           path="/var/infinispan/myapp/users">
            <write-behind flush-lock-timeout="15000" thread-pool-size="5" />
        </file-store>
    </persistence>
</local-cache>

In this example we enabled both eviction and a CacheStore (the persistence element). LIRS is one of the choices we have for eviction strategies. Here it is configured to keep (approximately) 2000 entries in live memory and evict the remaining as a memory usage control strategy.

The CacheStore is enabling passivation, which means that the entries which are evicted are stored on the filesystem.

You could configure an eviction strategy while not configuring a passivating CacheStore! That is a valid configuration for Infinispan but will have the evictor permanently remove entries. Hibernate OGM will break in such a configuration.

9.3.6. Clustering: store data on multiple Infinispan nodes

The best thing about Infinispan is that all nodes are treated equally and it requires almost no beforehand capacity planning: to add more nodes to the cluster you just have to start new JVMs, on the same or different physical servers, having your same Infinispan configuration and your same application.

Infinispan supports several clustering cache modes; each mode provides the same API and functionality but with different performance, scalability and availability options:

Infinispan cache modes
local

Useful for a single VM: networking stack is disabled

replication

All data is replicated to each node; each node contains a full copy of all entries. Consequentially reads are faster but writes don’t scale as well. Not suited for very large datasets.

distribution

Each entry is distributed on multiple nodes for redundancy and failure recovery, but not to all the nodes. Provides linear scalability for both write and read operations. distribution is the default mode.

To use the replication or distribution cache modes Infinispan will use JGroups to discover and connect to the other nodes.

In the default configuration, JGroups will attempt to autodetect peer nodes using a multicast socket; this works out of the box in the most network environments but will require some extra configuration in cloud environments (which often block multicast packets) or in case of strict firewalls. See the JGroups reference documentation, specifically look for Discovery Protocols to customize the detection of peer nodes.

Nowadays, the JVM defaults to use IPv6 network stack; this will work fine with JGroups, but only if you configured IPv6 correctly. It is often useful to force the JVM to use IPv4.

It is also important to let JGroups know which networking interface you want to use; it will bind to one interface by default, but if you have multiple network interfaces that might not be the one you expect.

Example 20. JVM properties to set for clustering
#192.168.122.1 is an example IPv4 address
-Djava.net.preferIPv4Stack=true -Djgroups.bind_addr=192.168.122.1

You don’t need to use IPv4: JGroups is compatible with IPv6 provided you have routing properly configured and valid addresses assigned.

The jgroups.bind_addr needs to match a placeholder name in your JGroups configuration in case you don’t use the default one.

The default configuration uses distribution as cache mode and uses the jgroups-tcp.xml configuration for JGroups, which is contained in the Infinispan jar as the default configuration for Infinispan users. Let’s see how to reconfigure this:

Example 21. Reconfiguring cache mode and override JGroups configuration
<?xml version="1.0" encoding="UTF-8"?>
<infinispan
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="urn:infinispan:config:9.1 http://www.infinispan.org/schemas/infinispan-config-9.1.xsd"
    xmlns="urn:infinispan:config:9.1">

    <jgroups>
        <stack-file name="custom-stack" path="my-jgroups-conf.xml" />
    </jgroups>

    <cache-container name="HibernateOGM" default-cache="DEFAULT">
        <transport stack="custom-stack" />

        <!-- *************************************** -->
        <!--     Default cache used as template      -->
        <!-- *************************************** -->
        <distributed-cache name="DEFAULT" mode="SYNC">
            <locking striping="false" acquire-timeout="10000"
                concurrency-level="500" write-skew="false" />
            <transaction mode="NON_DURABLE_XA" />
            <state-transfer enabled="true" timeout="480000"
                await-initial-transfer="true" />
        </distributed-cache>

        <!-- Override the cache mode: -->
        <replicated-cache name="User" mode="SYNC">
            <locking striping="false" acquire-timeout="10000"
                concurrency-level="500" write-skew="false" />
            <transaction mode="NON_DURABLE_XA" />
            <state-transfer enabled="true" timeout="480000"
                await-initial-transfer="true" />
        </replicated-cache>

        <distributed-cache name="Order" mode="SYNC">
            <locking striping="false" acquire-timeout="10000"
                concurrency-level="500" write-skew="false" />
            <transaction mode="NON_DURABLE_XA" />
            <state-transfer enabled="true" timeout="480000"
                await-initial-transfer="true" />
        </distributed-cache>

        <distributed-cache name="associations_User_Order" mode="SYNC">
            <locking striping="false" acquire-timeout="10000"
                concurrency-level="500" write-skew="false" />
            <transaction mode="NON_DURABLE_XA" />
            <state-transfer enabled="true" timeout="480000"
                await-initial-transfer="true" />
        </distributed-cache>

    </cache-container>

</infinispan>

In the example above we specify a custom JGroups configuration file and set the cache mode for the default cache to distribution; this is going to be inherited by the Order and the associations_User_Order caches. But for User we have chosen (for the sake of this example) to use replication.

Now that you have clustering configured, start the service on multiple nodes. Each node will need the same configuration and jars.

We have just shown how to override the clustering mode and the networking stack for the sake of completeness, but you don’t have to!

Start with the default configuration and see if that fits you. You can fine tune these setting when you are closer to going in production.

9.3.7. Storage principles

To describe things simply, each entity is stored under a single key. The value itself is a map containing the columns / values pair.

Each association from one entity instance to (a set of) another is stored under a single key. The value contains the navigational information to the (set of) entity.

Properties and built-in types

Each entity is represented by a map. Each property or more precisely column is represented by an entry in this map, the key being the column name.

Hibernate OGM support by default the following property types:

  • java.lang.String

  • java.lang.Character (or char primitive)

  • java.lang.Boolean (or boolean primitive); Optionally the annotations @Type(type = "true_false"), @Type(type = "yes_no") and @Type(type = "numeric_boolean") can be used to map boolean properties to the characters 'T'/'F', 'Y'/'N' or the int values 0/1, respectively.

  • java.lang.Byte (or byte primitive)

  • java.lang.Short (or short primitive)

  • java.lang.Integer (or integer primitive)

  • java.lang.Long (or long primitive)

  • java.lang.Integer (or integer primitive)

  • java.lang.Float (or float primitive)

  • java.lang.Double (or double primitive)

  • java.math.BigDecimal

  • java.math.BigInteger

  • java.util.Calendar

  • java.util.Date

  • java.util.UUID

  • java.util.URL

Hibernate OGM doesn’t store null values in Infinispan, setting a value to null is the same as removing the corresponding entry from Infinispan.

This can have consequences when it comes to queries on null value.

Identifiers

Entity identifiers are used to build the key in which the entity is stored in the cache.

The key is comprised of the following information:

  • the identifier column names

  • the identifier column values

  • the entity table (for the CACHE_PER_KIND strategy)

In CACHE_PER_TABLE, the table name is inferred from the cache name. In CACHE_PER_KIND, the table name is necessary to identify the entity in the generic cache.

Example 22. Define an identifier as a primitive type
@Entity
public class Bookmark {

    @Id
    private Long id;

    private String title;

    // getters, setters ...
}
Table 1. Content of the Bookmark cache in CACHE_PER_TABLE
KEY MAP ENTRIES

["id"], [42]

id

42

title

"Hibernate OGM documentation"

Table 2. Content of the ENTITIES cache in CACHE_PER_KIND
KEY MAP ENTRIES

"Bookmark", ["id"], [42]

id

42

title

"Hibernate OGM documentation"

Example 23. Define an identifier using @EmbeddedId
@Embeddable
public class NewsID implements Serializable {

    private String title;
    private String author;

    // getters, setters ...
}

@Entity
public class News {

    @EmbeddedId
    private NewsID newsId;
    private String content;

    // getters, setters ...
}
Table 3. Content of the News cache in CACHE_PER_TABLE
KEY MAP ENTRIES

[newsId.author, newsId.title], ["Guillaume", "How to use Hibernate OGM ?"]

newsId.author

"Guillaume"

newsId.title

"How to use Hibernate OGM ?"

content

"Simple, just like ORM but with a NoSQL database"

Table 4. Content of the ENTITIES cache in CACHE_PER_KIND
KEY MAP ENTRIES

"News", [newsId.author, newsId.title], ["Guillaume", "How to use Hibernate OGM ?"]

newsId.author

"Guillaume"

newsId.title

"How to use Hibernate OGM ?"

content

"Simple, just like ORM but with a NoSQL database"

Identifier generation strategies

Since Infinispan has not native sequence nor identity column support, these are simulated using the table strategy, however their default values vary. We highly recommend you explicitly use a TABLE strategy if you want to generate a monotonic identifier.

But if you can, use a pure in-memory and scalable strategy like a UUID generator.

Example 24. Id generation strategy TABLE using default values
@Entity
public class GuitarPlayer {

    @Id
    @GeneratedValue(strategy = GenerationType.TABLE)
    private long id;

    private String name;

    // getters, setters ...
}
Table 5. Content of the hibernate_sequences cache in CACHE_PER_TABLE
KEY NEXT VALUE

["sequence_name"], ["default"]

2

Table 6. Content of the IDENTIFIERS cache in CACHE_PER_KIND
KEY NEXT VALUE

"hibernate_sequences", ["sequence_name"], ["default"]

2

As you can see, in CACHE_PER_TABLE, the key does not contain the id source table name. It is inferred by the cache name hosting that key.

Example 25. Id generation strategy TABLE using a custom table
@Entity
public class GuitarPlayer {

    @Id
    @GeneratedValue(strategy = GenerationType.TABLE, generator = "guitarGen")
    @TableGenerator(
        name = "guitarGen",
        table = "GuitarPlayerSequence",
        pkColumnName = "seq"
        pkColumnValue = "guitarPlayer",
    )
    private long id;

    // getters, setters ...
}
Table 7. Content of the GuitarPlayerSequence cache in CACHE_PER_TABLE
KEY NEXT VALUE

["seq"], ["guitarPlayer"]

2

Table 8. Content of the IDENTIFIERS cache in CACHE_PER_KIND
KEY NEXT VALUE

"GuitarPlayerSequence", ["seq"], ["guitarPlayer"]

2

Example 26. SEQUENCE id generation strategy
@Entity
public class Song {

  @Id
  @GeneratedValue(strategy = GenerationType.SEQUENCE, generator = "songSequenceGenerator")
  @SequenceGenerator(
      name = "songSequenceGenerator",
      sequenceName = "song_sequence",
      initialValue = 2,
      allocationSize = 20
  )
  private Long id;

  private String title;

  // getters, setters ...
}
Table 9. Content of the hibernate_sequences cache in CACHE_PER_TABLE
KEY NEXT VALUE

["sequence_name"], ["song_sequence"]

11

Table 10. Content of the IDENTIFIERS cache in CACHE_PER_KIND
KEY NEXT VALUE

"hibernate_sequences", "["sequence_name"], ["song_sequence"]

11

Entities

Entities are stored in the cache named after the entity name when using the CACHE_PER_TABLE strategy. In the CACHE_PER_KIND strategy, entities are stored in a single cache named ENTITIES.

The key is comprised of the following information:

  • the identifier column names

  • the identifier column values

  • the entity table (for the CACHE_PER_KIND strategy)

In CACHE_PER_TABLE, the table name is inferred from the cache name. In CACHE_PER_KIND, the table name is necessary to identify the entity in the generic cache.

The entry value is an instance of org.infinispan.atomic.FineGrainedMap which contains all the entity properties - or to be specific columns. Each column name and value is stored as a key / value pair in the map. We use this specialized map as Infinispan is able to transport changes in a much more efficient way.

Example 27. Default JPA mapping for an entity
@Entity
public class News {

    @Id
    private String id;
    private String title;

    // getters, setters ...
}
Table 11. Content of the News cache in CACHE_PER_TYPE
KEY MAP ENTRIES

["id"], ["1234-5678"]

id

"1234-5678"

title

"On the merits of NoSQL"

Table 12. Content of the ENTITIES cache in CACHE_PER_KIND
KEY MAP ENTRIES

"News", ["id"], ["1234-5678"]

id

"1234-5678"

title

"On the merits of NoSQL"

As you can see, the table name is not part of the key for CACHE_PER_TYPE. In the rest of this section we will no longer show the CACHE_PER_KIND strategy.

Example 28. Rename field and collection using @Table and @Column
@Entity
@Table(name = "Article")
public class News {

    @Id
    private String id;

    @Column(name = "headline")
    private String title;

    // getters, setters ...
}
Table 13. Content of the Article cache
KEY MAP ENTRIES

["id"], ["1234-5678"]

id

"1234-5678"

headline

"On the merits of NoSQL"

Embedded objects and collections
Example 29. Embedded object
@Entity
public class News {

    @Id
    private String id;
    private String title;

    @Embedded
    private NewsPaper paper;

    // getters, setters ...
}

@Embeddable
public class NewsPaper {

    private String name;
    private String owner;

    // getters, setters ...
}
Table 14. Content of the News cache
KEY MAP ENTRIES

["id"], ["1234-5678"]

id

"1234-5678"

title

"On the merits of NoSQL"

paper.name

"NoSQL journal of prophecies"

paper.owner

"Delphy"

Example 30. @ElementCollection with one attribute
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
Table 15. Content of the GrandMother cache
KEY MAP ENTRIES

["id"], ["granny"]

id

"granny"

Table 16. Content of the associations_GrandMother_grandChildren cache in CACHE_PER_TYPE
KEY ROW KEY ROW MAP ENTRIES

["GrandMother_id"], ["granny"]

["GrandMother_id", "name"], ["granny", "Leia"]

GrandMother_id

"granny"

name

"Leia"

["GrandMother_id", "name"], ["granny", "Luke"]

GrandMother_id

"granny"

name

"Luke"

Table 17. Content of the ASSOCIATIONS cache in CACHE_PER_KIND
KEY ROW KEY ROW MAP ENTRIES

"GrandMother_grandChildren", ["GrandMother_id"], ["granny"]

["GrandMother_id", "name"], ["granny", "Leia"]

GrandMother_id

"granny"

name

"Leia"

["GrandMother_id", "name"], ["granny", "Luke"]

GrandMother_id

"granny"

name

"Luke"

Here, we see that the collection of elements is stored in a separate cache and entry. The association key is made of:

  • the foreign key column names pointing to the owner of this association

  • the foreign key column values pointing to the owner of this association

  • the association table name in the CACHE_PER_KIND approach where all associations share the same cache

The association entry is a map containing the representation of each entry in the collection. The keys of that map are made of:

  • the names of the columns uniquely identifying that specific collection entry (e.g. for a Set this is all of the columns)

  • the values of the columns uniquely identifying that specific collection entry

The value attack to that collection entry key is a Map containing the key value pairs column name / column value.

Example 31. @ElementCollection with @OrderColumn
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    @OrderColumn( name = "birth_order" )
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
Table 18. Content of the GrandMother cache
KEY MAP ENTRIES

["id"], ["granny"]

id

"granny"

Table 19. Content of the GrandMother_grandChildren cache
KEY ROW KEY ROW MAP ENTRIES

["GrandMother_id"], ["granny"]

["GrandMother_id", "birth_order"], ["granny", 0]

GrandMother_id

"granny"

birth_order

0

name

"Leia"

["GrandMother_id", "birth_order"], ["granny", 1]

GrandMother_id

"granny"

birth_order

1

name

"Luke"

Here we used an indexed collection and to identify the entry in the collection, only the owning entity id and the index value is enough.

Example 32. @ElementCollection with Map of @Embeddable
@Entity
public class ForumUser {

    @Id
    private String name;

    @ElementCollection
    private Map<String, JiraIssue> issues = new HashMap<>();

    // getters, setters ...
}

@Embeddable
public class JiraIssue {

    private Integer number;
    private String project;

    // getters, setters ...
}
Table 20. Content of the ForumUser cache
KEY MAP ENTRIES

["id"], ["Jane Doe"]

id

"Jane Doe"

Table 21. Content of the ForumUser_issues cache
KEY ROW KEY ROW MAP ENTRIES

["ForumUser_id"], ["Jane Doe"]

["ForumUser_id", "issues_KEY"], ["Jane Doe", "issueWithNull"]

ForumUser_id

Jane Doe

issue_KEY

"issueWithNull"

issues.value.project

<null>

issues.value.number

<null>

["ForumUser_id", "issues_KEY"], ["Jane Doe", "issue1"]

ForumUser_id

"Jane Doe"

issue_KEY

"issue1"

issues.value.project

"OGM"

issues.value.number

1253

["ForumUser_id", "issues_KEY"], ["Jane Doe", "issue2"]

ForumUser_id

"Jane Doe"

issue_KEY

"issue2"

issues.value.project

"HSEARCH"

issues.value.number

2000

Associations

Associations between entities are mapped like (collection of) embeddables except that the target entity is represented by its identifier(s).

Example 33. Unidirectional one-to-one
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}

@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    private Vehicule vehicule;

    // getters, setters ...
}
Table 22. Content of the Vehicule cache
KEY MAP ENTRIES

["id"], ["V_01"]

id

"V_01"

brand

"Mercedes"

Table 23. Content of the Wheel cache
KEY MAP ENTRIES

["id"], ["W001"]

id

"W001"

diameter

0.0

vehicule_id

"V_01"

Example 34. Unidirectional one-to-one with @JoinColumn
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}


@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    @JoinColumn( name = "part_of" )
    private Vehicule vehicule;

    // getters, setters ...
}
Table 24. Content of the Vehicle cache
KEY MAP ENTRIES

["id"], ["V_01"]

id

"V_01"

brand

"Mercedes"

Table 25. Content of the Wheel cache
KEY MAP ENTRIES

"Wheel", ["id"], ["W001"]

id

"W001"

diameter

0.0

part_of

"V_01"

Example 35. Unidirectional one-to-one with @MapsId and @PrimaryKeyJoinColumn
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}

@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    @PrimaryKeyJoinColumn
    @MapsId
    private Vehicule vehicule;

    // getters, setters ...
}
Table 26. Content of the Vehicle cache
KEY MAP ENTRIES

["id"], ["V_01"]

id

"V_01"

brand

"Mercedes"

Table 27. Content of the Wheel cache
KEY MAP ENTRIES

["vehicule_id"], ["V_01"]

vehicule_id

"V_01"

diameter

0.0

Example 36. Bidirectional one-to-one
@Entity
public class Husband {

    @Id
    private String id;
    private String name;

    @OneToOne
    private Wife wife;

    // getters, setters ...
}

@Entity
public class Wife {

    @Id
    private String id;
    private String name;

    @OneToOne(mappedBy="wife")
    private Husband husband;

    // getters, setters ...
}
Table 28. Content of the Husband cache
KEY MAP ENTRIES

["id"], ["alex"]

id

"alex"

name

"Alex"

wife

"bea"

Table 29. Content of the Wife cache
KEY MAP ENTRIES

["id"], ["bea"]

id

"bea"

name

"Bea"

Table 30. Content of the associations_Husband cache
KEY ROW KEY MAP ENTRIES

["wife"], ["bea"]

["id", "wife"], ["alex", "bea"]

id

"alex"

wife

"bea"

Example 37. Unidirectional one-to-many
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}
Table 31. Content of the Basket cache
KEY MAP ENTRIES

["id"], ["davide_basket"]

id

"davide_basket"

owner

"Davide"

Table 32. Content of the Product cache
KEY MAP ENTRIES

["name"], ["Beer"]

name

"Beer"

description

"Tactical Nuclear Penguin"

["name"], ["Pretzel"]

name

"Pretzel"

description

"Glutino Pretzel Sticks"

Table 33. Content of the associations_Basket_Product cache
KEY ROW KEY MAP ENTRIES

["Basket_id"], ["davide_basket"]

["Basket_id", "products_name"], ["davide_basket", "Beer"]

Basket_id

"davide_basket"

products_name

"Beer"

["Basket_id", "products_name"], ["davide_basket", "Pretzel"]

Basket_id

"davide_basket"

products_name

"Pretzel"

Example 38. Unidirectional one-to-many with @JoinTable
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    @JoinTable( name = "BasketContent" )
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}
Table 34. Content of the Basket cache
KEY MAP ENTRIES

["id"], ["davide_basket"]

id

"davide_basket"

owner

"Davide"

Table 35. Content of the Basket cache
KEY MAP ENTRIES

["name"], ["Beer"]

name

"Beer"

description

"Tactical Nuclear Penguin"

["name"], ["Pretzel"]

name

"Pretzel"

description

"Glutino Pretzel Sticks"

Table 36. Content of the associations_BasketContent cache
KEY ROW KEY MAP ENTRIES

["Basket_id"], ["davide_basket"]

["Basket_id", "products_name"], ["davide_basket", "Beer"]

Basket_id

"davide_basket"

products_name

"Beer"

["Basket_id", "products_name"], ["davide_basket", "Pretzel"]

Basket_id

"davide_basket"

products_name

"Pretzel"

Example 39. Unidirectional one-to-many using maps with defaults
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    private Map<String, Address> addresses = new HashMap<String, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}
Table 37. Content of the User cache
KEY MAP ENTRIES

["id"], ["user_001"]

id

"user_001"

Table 38. Content of the Address cache
KEY MAP ENTRIES

["id"], ["address_001"]

id

"address_001"

city

"Rome"

["id"], ["address_002"]

id

"address_002"

city

"Paris"

Table 39. Content of the associations_User_address cache
KEY ROW KEY MAP ENTRIES

["User_id"], "user_001"]

["User_id", "addresses_KEY"], ["user_001", "home"]

User_id

"user_001"

addresses_KEY

"home"

addresses_id

"address_001"

["User_id", "addresses_KEY"], ["user_001", "work"]

User_id

"user_002"

addresses_KEY

"work"

addresses_id

"address_002"

Example 40. Unidirectional one-to-many using maps with @MapKeyColumn
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    @MapKeyColumn(name = "addressType")
    private Map<String, Address> addresses = new HashMap<String, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}
Table 40. Content of the User cache
KEY MAP ENTRIES

["id"], ["user_001"]

id

"user_001"

Table 41. Content of the Address cache
KEY MAP ENTRIES

["id"], ["address_001"]

id

"address_001"

city

"Rome"

["id"], ["address_002"]

id

"address_002"

city

"Paris"

Table 42. Content of the associations_User_address cache
KEY ROW KEY MAP ENTRIES

["User_id"], "user_001"]

["User_id", "addressType"], ["user_001", "home"]

User_id

"user_001"

addressesType

"home"

addresses_id

"address_001"

["User_id", "addressType"], ["user_001", "work"]

User_id

"user_002"

addressesType

"work"

addresses_id

"address_002"

Example 41. Unidirectional many-to-one
@Entity
public class JavaUserGroup {

    @Id
    private String jugId;
    private String name;

    // getters, setters ...
}

@Entity
public class Member {

    @Id
    private String id;
    private String name;

    @ManyToOne
    private JavaUserGroup memberOf;

    // getters, setters ...
}
Table 43. Content of the JavaUserGroup cache
KEY MAP ENTRIES

["jugId"], ["summer_camp"]

jugId

"summer_camp"

name

"JUG Summer Camp"

Table 44. Content of the Member cache
KEY MAP ENTRIES

["member_id"], ["emmanuel"]

member_id

"emmanuel"

name

"Emmanuel Bernard"

memberOf_jug_id

"summer_camp"

["member_id"], ["jerome"]

member_id

"jerome"

name

"Jerome"

memberOf_jug_id

"summer_camp"

Example 42. Bidirectional many-to-one
@Entity
public class SalesForce {

    @Id
    private String id;
    private String corporation;

    @OneToMany(mappedBy = "salesForce")
    private Set<SalesGuy> salesGuys = new HashSet<SalesGuy>();

    // getters, setters ...
}

@Entity
public class SalesGuy {
    private String id;
    private String name;

    @ManyToOne
    private SalesForce salesForce;

    // getters, setters ...
}
Table 45. Content of the SalesForce cache
KEY MAP ENTRIES

["id"], ["red_hat"]

id

"red_hat"

corporation

"Red Hat"

Table 46. Content of the SalesGuy cache
KEY MAP ENTRIES

["id"], ["eric"]

id

"eric"

name

"Eric"

salesForce_id

"red_hat"

["id"], ["simon"]

id

"simon"

name

"Simon"

salesForce_id

"red_hat"

Table 47. Content of the associations_SalesGuy cache
KEY ROW KEY MAP ENTRIES

["salesForce_id"], ["red_hat"]

["salesForce_id", "id"], ["red_hat", "eric"]

salesForce_id

"red_hat"

id

"eric"

["salesForce_id", "id"], ["red_hat", "simon"]

salesForce_id

"red_hat"

id

"simon"

Example 43. Unidirectional many-to-many
@Entity
public class Student {

    @Id
    private String id;
    private String name;

    // getters, setters ...
}

@Entity
public class ClassRoom {

    @Id
    private long id;
    private String lesson;

    @ManyToMany
    private List<Student> students = new ArrayList<Student>();

    // getters, setters ...
}

The "Math" class has 2 students: John Doe and Mario Rossi

The "English" class has 2 students: Kate Doe and Mario Rossi

Table 48. Content of the ClassRoom cache
KEY MAP ENTRIES

["id"], [1]

id

1

name

"Math"

["id"], [2]

id

2

name

"English"

Table 49. Content of the Student cache
KEY MAP ENTRIES

["id"], ["john"]

id

"john"

name

"John Doe"

["id"], ["mario"]

id

"mario"

name

"Mario Rossi"

["id"], ["kate"]

id

"kate"

name

"Kate Doe"

Table 50. Content of the associations_ClassRoom_Student cache
KEY ROW KEY MAP ENTRIES

["ClassRoom_id"], [1]

["ClassRoom_id", "students_id"], [1, "mario"]

ClassRoom_id

1

students_id

"mario"

["ClassRoom_id", "students_id"], [1, "john"]

ClassRoom_id

1

students_id

"john"

["ClassRoom_id"], [2]

["ClassRoom_id", "students_id"], [2, "kate"]

ClassRoom_id

2

students_id

"kate"

["ClassRoom_id", "students_id"], [2, "mario"]

ClassRoom_id

2

students_id

"mario"

Example 44. Bidirectional many-to-many
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany( mappedBy = "bankAccounts" )
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}

David owns 2 accounts: "012345" and "ZZZ-009"

Table 51. Content of the AccountOwner cache
KEY MAP ENTRIES

["id"], ["David"]

id

"David"

SSN

"0123456"

Table 52. Content of the BankAccount cache
KEY MAP ENTRIES

["id"], ["account_1"]

id

"account_1"

accountNumber

"X2345000"

["id"], ["account_2"]

id

"account_2"

accountNumber

"ZZZ-009"

Table 53. Content of the AccountOwner_BankAccount cache
KEY ROW KEY MAP ENTRIES

["bankAccounts_id"], ["account_1"]

["bankAccounts_id", "owners_id"], ["account_1", "David"]

bankAccounts_id

"account_1"

owners_id

"David"

["bankAccounts_id"], ["account_2"]

["bankAccounts_id", "owners_id"], ["account_2", "David"]

bankAccounts_id

"account_2"

owners_id

"David"

["owners_id"], ["David"]

["owners_id", "banksAccounts_id"], ["David", "account_1"]

bankAccounts_id

"account_1"

owners_id

"David"

["owners_id", "banksAccounts_id"], ["David", "account_2"]

bankAccounts_id

"account_2"

owners_id

"David"

9.3.8. Transactions

Infinispan supports transactions and integrates with any standard JTA TransactionManager; this is a great advantage for JPA users as it allows to experience a similar behaviour to the one we are used to when we work with RDBMS databases.

This capability is only available to Infinispan Embedded users: the transactional integration capabilities are not exposed to the Hot Rod clients.

If you’re having Hibernate OGM start and manage Infinispan, you can skip this as it will inject the same TransactionManager instance which you already have set up in the Hibernate / JPA configuration.

If you are providing an already started Infinispan CacheManager instance by using the JNDI lookup approach, then you have to make sure the CacheManager is using the same TransactionManager as Hibernate:

Example 45. Configuring a JBoss Standalone TransactionManager lookup in Infinispan configuration
<default>
   <transaction
      transactionMode="TRANSACTIONAL"
      transactionManagerLookupClass=
    "org.infinispan.transaction.lookup.JBossStandaloneJTAManagerLookup" />
</default>

Infinispan supports different transaction modes like PESSIMISTIC and OPTIMISTIC, supports XA recovery and provides many more configuration options; see the Infinispan User Guide for more advanced configuration options.

9.3.9. Storing a Lucene index in Infinispan

Hibernate Search, which can be used for advanced query capabilities (see Query your entities), needs some place to store the indexes for its embedded Apache Lucene engine.

A common place to store these indexes is the filesystem which is the default for Hibernate Search; however if your goal is to scale your NoSQL engine on multiple nodes you need to share this index. Network sharing file systems are a possibility but we don’t recommended that. Often the best option is to store the index in whatever NoSQL database you are using (or a different dedicated one).

You might find this section useful even if you don’t intend to store your data in Infinispan.

The Infinispan project provides an adaptor to plug into Apache Lucene, so that it writes the indexes in Infinispan and searches data in it. Since Infinispan can be used as an application cache to other NoSQL storage engines by using a CacheStore (see Manage data size) you can use this adaptor to store the Lucene indexes in any NoSQL store supported by Infinispan:

  • JDBC databases

  • Cassandra

  • Filesystem (but locked correctly at the Infinispan level)

  • MongoDB

  • HBase

  • LevelDB

  • A secondary (independent) Infinispan grid

How to configure it? Here is a simple cheat sheet to get you started with this type of setup:

  • Add org.infinispan:infinispan-directory-provider:{infinispanVersion} to your dependencies

  • set these configuration properties:

    • hibernate.search.default.directory_provider = infinispan

    • hibernate.search.default.exclusive_index_use = false

    • hibernate.search.infinispan.configuration_resourcename = [infinispan configuration filename]

This configuration is simple and will work fine in most scenarios, but keep in mind that using 'exclusive_index_use' will be neither fast nor scalable. For high performance, high concurrency or production use please refer to the Infinispan documentation for more advanced configuration options and tuning.

The referenced Infinispan configuration should define a CacheStore to load/store the index in the NoSQL engine of choice. It should also define three cache names:

Table 54. Infinispan caches used to store indexes
Cache name Description Suggested cluster mode

LuceneIndexesLocking

Transfers locking information. Does not need a cache store.

replication

LuceneIndexesData

Contains the bulk of Lucene data. Needs a cache store.

distribution + L1

LuceneIndexesMetadata

Stores metadata on the index segments. Needs a cache store.

replication

This configuration is not going to scale well on write operations: to do that you should read about the master/slave and sharding options in Hibernate Search. The complete explanation and configuration options can be found in the Hibernate Search Reference Guide

Some NoSQL support storage of Lucene indexes directly, in which case you might skip the Infinispan Lucene integration by implementing a custom DirectoryProvider for Hibernate Search. You’re very welcome to share the code and have it merged in Hibernate Search for others to use, inspect, improve and maintain.

9.4. Hibernate OGM & Infinispan Server over Hot Rod

In this section we’ll see how to configure Hibernate OGM to connect to "Infinispan Server using the Hot Rod protocol", which we will call "Infinispan Remote" for brevity and to differentiate it from "Infinispan Embedded".

In this mode Hibernate OGM can not boostrap or otherwise control the lifecycle of Infinispan, so we will assume that you already have a cluster of Infinispan Server nodes running. For instructions on setting one up, see the Infinispan Server Guide.

The good news is that - since it’s a separate service - there won’t be much to configure in Hibernate OGM.

The Hibernate OGM support for Infinispan Remote is considered experimental. In particular, the storage format is not set in stone.

9.4.1. Adding Infinispan Remote dependencies

To use Hibernate OGM to connect to an Infinispan Server using the Hot Rod protocol, you will need the following extension and its transitive dependencies (which include, among others, the Hot Rod client):

<dependency>
    <groupId>org.hibernate.ogm</groupId>
    <artifactId>hibernate-ogm-infinispan-remote</artifactId>
    <version>5.2.0.Beta1</version>
</dependency>

9.4.2. Configuration properties for Infinispan Remote

First, let Hibernate know that you want to use the OGM Infinispan Remote datastore by setting the hibernate.ogm.datastore.provider property to infinispan_remote.

The next step is to configure the Hot Rod client. You have two options:

  • either provide a resource file containing all Hot Rod client configuration properties

  • or include all the Hot Rod client configuration properties with a custom prefix, as explained below.

To use an external configuration resource, set the hibernate.ogm.infinispan_remote.configuration_resource_name configuration property to the resource name.

Example 46. Using a separate resource to configure the Hot Rod client
<?xml version="1.0"?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_2_0.xsd"
    version="2.0">

    <persistence-unit name="ogm-with-hotrod">
        <provider>org.hibernate.ogm.jpa.HibernateOgmPersistence</provider> (1)
        <properties>
            <property name="hibernate.ogm.datastore.provider"
                value="infinispan_remote" /> (2)
            <property name="hibernate.ogm.infinispan_remote.configuration_resource_name"
                value="hotrodclient.properties" /> (3)
        </properties>
    </persistence-unit>
</persistence>
1 Choose Hibernate OGM as JPA Provider
2 pick infinispan_remote as datastore
3 point to the Hot Rod configuration file
infinispan.client.hotrod.server_list = 127.0.0.1:11222
infinispan.client.hotrod.tcp_no_delay = true
infinispan.client.hotrod.tcp_keep_alive = false

## below is connection pooling config
maxActive=-1
maxTotal = -1
maxIdle = -1
whenExhaustedAction = 1
timeBetweenEvictionRunsMillis = 120000
minEvictableIdleTimeMillis = 300000
testWhileIdle = true
minIdle = 1

Alternatively you can embed the Hot Rod properties in your Hibernate (or JPA) configuration file, but you’ll have to replace the infinispan.client.hotrod. prefix with the custom prefix hibernate.ogm.infinispan_remote.client..

Some of the Hot Rod client configuration properties don’t normally use a prefix - specifically all properties relating to connection pooling as in the previous example - these will also need to use the hibernate.ogm.infinispan_remote.client. prefix.

Properties set with the hibernate.ogm.infinispan_remote.client. prefix will override the same properties configured using an external reosurce file.

Example 47. Embedding the Hot Rod client configuration properties in the Hibernate configuration
<?xml version="1.0"?>
<persistence xmlns="http://java.sun.com/xml/ns/persistence"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_2_0.xsd"
    version="2.0">

    <persistence-unit name="ogm-with-hotrod">
        <provider>org.hibernate.ogm.jpa.HibernateOgmPersistence</provider> (1)
        <properties>
            <property name="hibernate.ogm.datastore.provider"
                value="infinispan_remote" /> (2)
            <property name="hibernate.ogm.infinispan_remote.client.server_list"
                value="127.0.0.1:11222" /> (3)
            <property name="hibernate.ogm.infinispan_remote.client.tcp_no_delay"
                value="true" />
            <property name="hibernate.ogm.infinispan_remote.client.tcp_keep_alive"
                value="false" />
            <property name="hibernate.ogm.infinispan_remote.client.maxActive"
                value="-1" />
            <property name="hibernate.ogm.infinispan_remote.client.maxTotal"
                value="-1" />
            <property name="hibernate.ogm.infinispan_remote.client.maxIdle"
                value="-1" />
            <property name="hibernate.ogm.infinispan_remote.client.whenExhaustedAction"
                value="1" />
            <property name="hibernate.ogm.infinispan_remote.client.timeBetweenEvictionRunsMillis"
                value="120000" />
            <property name="hibernate.ogm.infinispan_remote.client.minEvictableIdleTimeMillis"
                value="300000" />
            <property name="hibernate.ogm.infinispan_remote.client.testWhileIdle"
                value="true" />
            <property name="hibernate.ogm.infinispan_remote.client.minIdle"
                value="1" />
        </properties>
    </persistence-unit>
</persistence>
1 Choose Hibernate OGM as JPA Provider
2 pick infinispan_remote as datastore
3 include Hot Rod configuration properties, just replacing/adding the OGM prefix.

In the next section we’ll see a couple more advanced properties which might be of interest.

hibernate.ogm.infinispan_remote.schema_capture_service

If you set this to an implementation of org.hibernate.ogm.datastore.infinispanremote.schema.spi.SchemaCapture you can collect any generated Protobuf Schema. Could be useful for integrations with other tools. You can either provide a fully qualified classname or a SchemaCapture, or pass an instance of a SchemaCapture in the configuration properties, if you’re booting Hibernate programmatically.

hibernate.ogm.infinispan_remote.schema_package_name

Defines the package name of the generated Protobuf schema. Defaults to HibernateOGMGenerated. Useful to isolate different applications using the same Infinispan Server instance.

9.4.3. Data encoding: Protobuf Schema

Using the Infinispan Remote backend your data will be encoded using Protocol Buffers, also known as Protobuf.

Protocol Buffers are a language-neutral, platform-neutral extensible mechanism for serializing structured data

— https://developers.google.com/protocol-buffers/

This encoding strategy will be used both during transmission to and from the datagrid, and as a storage format on the Infinispan Server.

Typical usage of Google’s developer tools for Java would require you to download the protoc compiler to generate Java stubs; you won’t need that when using Hibernate OGM as the backend will generate the encoding and decoding functions on the fly from your entities.

The benefit of having Hibernate OGM generate the schema for you will make it easier to get started, but there’s a drawback: you are not directly in control of the protobuf schema It will deploy this schema - or expect a compatible schema to be deployed - as it will use its generated codecs to read and write data to the Infinispan Server.

The protobuf technology is designed to allow evolution of your schema: you can deploy a different schema on the Infinispan Server than the one OGM expects, but this is an advanced topic and you’ll have to make sure the deployed schema is compatible with the one OGM is generating and using.

Another reason to make sure the deployed protobuf schema is a compatible evolution of a previous schema, is to make sure you can still read data which is already stored in the datagrid.

Remember that the Protobuf schema is used both during transmission and storage. The fact that it’s used also during transmission of your data is a key difference to the schema of a SQL database.

For example even if a property "A" is not nullable in terms of storage, you will still want it to be flagged as optional in a protobuf schema to allow, for example, retrieving a subset of data properties without having to always retrieve the property "A".

You don’t need to do anything regarding the schema: Hibernate OGM will automatically deploy it to the Infinispan datagrid at boostrap of Hibernate. You might want to keep this in mind though, both to be able to evolve your schema without data loss, and to be able to generate decoders for other Infinispan clients not using Hibernate OGM.

The deployed schemas can be fetched from the Infinispan Server; Hibernate OGM also logs the generated schemas at INFO level in the logging category org.hibernate.ogm.datastore.infinispanremote.impl.protobuf.SchemaDefinitions.

9.4.4. Storage Principles of the Infinispan Remote dataprovider

This is actually very simple.

Imagine you were mapping your entities to a traditional, table based RDBMS; now instead of tables, you have caches. Each cache has a name, and a consistent schema, and for each cache we define a key with some properties (the id, aka the primary key).

Relations are mapped by encoding a "foreign key"; these are used either as keys perform a key lookup on another table, or can be used in queries on other tables to identify relations which have a higher than one cardinality.

So let’s highlight the differences with the relational world:

Referential integrity

While we can use relations based on foreign keys, Infinispan has no notion of referential integrity. Hibernate is able to maintain the integrity as it won’t "forget" stale references, but since the storage doesn’t support transactions either it is possible to interrupt Hibernate OGM during such maintenance and introduce breaks of integrity.

A key. And a Value.

In a key/value store the two elements key and value are different, separate objects. The schema - and consequentially all operations - generated by Hibernate OGM will treat and encode these two objects separately. You will notice that the attributes of the key are encoded in the value as well, as it is not possible to run e.g. range queries on attributes of keys.

No Sequences, no auto-incrementing values

Infinispan does not support sequences, yet allows concurrent "compare and set" operations; Hibernate OGM makes use of such CAS operations to emulate the need of sequences or auto-incrementing primary keys if your entity mapping uses them, however this solution might not work under high load: make sure to use a different strategy, such as assigning IDs explicitly, or using the org.hibernate.id.UUIDGenerator generator. Hibernate OGM will log a warning if it detects excessive spinning on such CAS operations.

Not mapped to JDBC types, but to Protobuf types

Rather than mapping your Java properties to corresponding JDBC (SQL) types, your Java properties are mapped to Protobuf types. See the protobuf documentation for an overview of protocol buffer "primitive" types.

Example 48. Example auto-generated Protobuf Schema for a simple entity
import javax.persistence.Column;
import javax.persistence.Entity;
import javax.persistence.Id;

@Entity
public class Hypothesis {

    @Id String id;

    String description;

    @Column(name = "pos")
    int position;

}
package HibernateOGMGenerated; (1)

message Hypothesis_id { (2)
    required string id = 1;
}

message Hypothesis {
    required string id = 1;
    optional string description = 2;
    optional int32 pos = 3;  (3)
}
1 The default Protobuf package name.
2 A dedicated message type for the Key of the Key/Value pair
3 The pos attribute name respects the option of the @Column annotation

The above example shows how a Protobuf schema looks like, as automatically generated from a mapped entity. Any property type supported by Hibernate ORM will be converted to a matching Protobuf type.

Each Table requires a Cache with the same name

In a relational database world, when Hibernate defines the schema this implicitly creates the tables; this is not the case on Infinispan.

With Infinispan, the Protobuf Schema just unlocks the capability to transmit messages with such payloads (read/write), and allows the remote servers to process the fields, for example to execute queries and extract projections out of the stored entries. So this establishes a transmission and storage encoding contract, but doesn’t actually start or allocate any storing Cache.

Hibernate OGM by convention will write to several named Caches, mapping each "table name" to a "cache name". In the above example, when having an Hypothesis entity this will write to a Cache named Hypothesis.

The benefit is that you can tune each cache (each "table") independently; for example you could configure the caches for the most important data to have a synchronous CacheStore which replicates data to a relational database, and have less important entries use an asynchronous CacheStore, or none at all, to favour performance over redundancy.

The drawback of this design choice is that each named cache must be pre-defined in the Infinispan Server configuration: at this point, the Hot Rod protocol is not allowed to start missing caches so Hibernate OGM can not define the missing tables automatically. It generates the encoding protobuf, but you have to list the Cache names in the server configuration.

For each "table name" your model would generate on a relational database, you have to define a matching Cache on the Infinispan Server.

If any Cache is missing, Hibernate OGM will fail to start and list which table names were expected, but not found. We plan to automate the creation of missing caches in the future.

9.4.5. Known Limitations & Future improvements

The Infinispan Remote dataprovider has some known limitations, some of which are unsolvable without further development of Infinispan itself.

Transaction Support

We’re eagerly waiting for Infinispan to support transactions over Hot Rod, as it already provides this feature in Embedded Mode.

Queries

At this point the Hibernate OGM backend is able to run the queries it needs to materialize relations, but does not yet translate JPQL queries nor Criteria queries to the Infinispan remote queries.

Indexing

Infinispan supports Hibernate Search annotations directly embedded within its protobuf schema definitions; this would enable the queries on them to use indexes. Hibernate OGM doesn’t generate these annotations in the schemas it generates yet.

Native support for write skew checks

The Hot Rod client has native support for versioning of datagrid entries, yet this is not supported on all of the client APIs. For Hibernate OGM to be able to consistently use versioning requires enhancements to the Hot Rod client API.

Enums

Protobuf has native support for Enum types, yet the JPA annotations force to choose between ordinal or string encoding. We might have to introduce a "native" encoding, probably via a novel mapping annotation. Hibernate OGM supports the native protobuf Encoding but the JPA metadata will always force the ordinal or string representations.

Nesting and embedding

The Protobuf schema could allow us to embed objects, including series of objects, as nested elements. This could allow mappings similar to the document based NoSQL stores, such as our MongoDB dialect, but is not supported yet.

Automatic creating of Caches

When deploying the Protobuf Schema, we should also automatically define and start the needed Caches if they are not defined. This is currently not allowed over the Hot Rod protocol.

10. MongoDB

MongoDB is a document oriented datastore written in C++ with strong emphasis on ease of use. The nested nature of documents make it a particularly natural fit for most object representations.

This implementation is based upon the MongoDB Java driver. The currently supported version is 3.4.

10.1. Why should I use Hibernate OGM with MongoDB

It is possible that in your project you have some entities that might benefit from MongoDB dynamic schema, but having a schema makes it possible to obtain better performance because the datastore can use the information of the schema to apply some optimizations that wouldn’t be otherwise possible.

JPA already has ways to define constraints and indexes and via Hibernate OGM you can use the same annotations for both your relational and not relational needs.

Hibernate OGM cannot make MongoDB transactional but by using the JPA transaction demarcation mechanism it can group the operations and flush them to the datastore to minimize the number of requests.

Another benefit of using Hibernate OGM with MongoDB is that it will also make it possible to use Hibernate Search out of the box. Hibernate Search brings the power of Lucene to your project, giving you the ability to run fast google-like searches.

This means that you can query the datastore using:

One of Hibernate OGM main goal is to map entities in a "natural" way, this means that your datastore will still be accessible in case you need to use other tools or want to run native queries occasionally.

10.2. Configuring MongoDB

Configuring Hibernate OGM to use MongoDb is easy:

  • Add the MongoDB module and driver to the classpath

  • provide the MongoDB URL to Hibernate OGM

10.2.1. Adding MongoDB dependencies

To add the dependencies via Maven, add the following module:

<dependency>
    <groupId>org.hibernate.ogm</groupId>
    <artifactId>hibernate-ogm-mongodb</artifactId>
    <version>5.2.0.Beta1</version>
</dependency>

This will pull the MongoDB driver transparently.

If you’re not using a dependency management tool, copy all the dependencies from the distribution in the directories:

  • /lib/required

  • /lib/mongodb

  • Optionally - depending on your container - you might need some of the jars from /lib/provided

MongoDB does not require Hibernate Search for the execution of JPQL or HQL queries.

10.2.2. MongoDB specific configuration properties

To get started quickly, pay attention to the following options:

  • hibernate.ogm.datastore.provider

  • hibernate.ogm.datastore.host

  • hibernate.ogm.datastore.database

And we should have you running. The following properties are available to configure MongoDB support:

MongoDB datastore configuration properties
hibernate.ogm.datastore.provider

To use MongoDB as a datastore provider, this property must be set to mongodb

hibernate.ogm.option.configurator

The fully-qualified class name or an instance of a programmatic option configurator (see Programmatic configuration)

hibernate.ogm.datastore.host

The hostname and port of the MongoDB instance. The optional port is concatenated to the host and separated by a colon. When using replica sets, you can define the various servers in a comma separated list of hosts and ports. Let’s see a few valid examples:

  • mongodb.example.com

  • mongodb.example.com:27018

  • 2001:db8::ff00:42:8329 (IPv6)

  • [2001:db8::ff00:42:8329]:27018 (IPv6 with port requires the IPv6 to be surrounded by square brackets)

  • www.example.com, www2.example.com:123, 192.0.2.1, 192.0.2.2:123, 2001:db8::ff00:42:8329, [2001:db8::ff00:42:8329]:123 (replica set)

    The default value is 127.0.0.1:27017. If left undefined, the default port is 27017.

hibernate.ogm.datastore.port

Deprecated: use hibernate.ogm.datastore.host. The port used by the MongoDB instance. Ignored when multiple hosts are defined. The default value is 27017.

hibernate.ogm.datastore.database

The database to connect to. This property has no default value.

hibernate.ogm.datastore.create_database

If set to true, the database will be created if it doesn’t exist. This property default value is false.

hibernate.ogm.datastore.username

The username used when connecting to the MongoDB server. This property has no default value.

hibernate.ogm.datastore.password

The password used to connect to the MongoDB server. This property has no default value. This property is ignored if the username isn’t specified.

hibernate.ogm.error_handler

The fully-qualified class name, class object or an instance of ErrorHandler to get notified upon errors during flushes (see Acting upon errors during application of changes)

hibernate.ogm.mongodb.driver.*

Defines a prefix for all options which should be passed through to the MongoDB driver. For available options refer to the JavaDocs of MongoClientOptions.Builder. All String, int and boolean properties can be set, eg hibernate.ogm.mongodb.driver.serverSelectionTimeout.

hibernate.ogm.mongodb.authentication_database

Defines the name of the authentication database, default value is admin.

hibernate.ogm.mongodb.authentication_mechanism

Defines the authentication mechanism to use. Possible values are:

  • BEST: Handshakes with the server to find the best authentication mechanism.

  • SCRAM_SHA_1: The SCRAM SHA 1 Challenge Response mechanism as described in this RFC.

  • MONGODB_CR: The MongoDB Challenge Response mechanism (deprecated since MongoDB 3)

  • GSSAPI: The GSSAPI mechanism. See the RFC

  • MONGODB_X509: The MongoDB X.509

  • PLAIN: The PLAIN mechanism. See the RFC

hibernate.ogm.datastore.document.association_storage

Defines the way OGM stores association information in MongoDB. The following two strategies exist (values of the org.hibernate.ogm.datastore.document.options.AssociationStorageType enum):

  • IN_ENTITY: store association information within the entity

  • ASSOCIATION_DOCUMENT: store association information in a dedicated document per association

IN_ENTITY is the default and recommended option unless the association navigation data is much bigger than the core of the document and leads to performance degradation.

hibernate.ogm.mongodb.association_document_storage

Defines how to store assocation documents (applies only if the ASSOCIATION_DOCUMENT association storage strategy is used). Possible strategies are (values of the org.hibernate.ogm.datastore.mongodb.options.AssociationDocumentStorageType enum):

  • GLOBAL_COLLECTION (default): stores the association information in a unique MongoDB collection for all associations

  • COLLECTION_PER_ASSOCIATION stores the association in a dedicated MongoDB collection per association

hibernate.ogm.datastore.document.map_storage

Defines the way OGM stores the contents of map-typed associations in MongoDB. The following two strategies exist (values of the org.hibernate.ogm.datastore.document.options.MapStorageType enum):

  • BY_KEY: map-typed associations with a single key column which is of type String will be stored as a sub-document, organized by the given key; Not applicable for other types of key columns, in which case always AS_LIST will be used

  • AS_LIST: map-typed associations will be stored as an array containing a sub-document for each map entry. All key and value columns will be contained within the array elements

hibernate.ogm.mongodb.write_concern

Defines the write concern setting to be applied when issuing writes against the MongoDB datastore. Possible settings are (values of the WriteConcernType enum): ACKNOWLEDGED, UNACKNOWLEDGED, FSYNCED, JOURNALED, REPLICA_ACKNOWLEDGED, MAJORITY and CUSTOM. When set to CUSTOM, a custom WriteConcern implementation type has to be specified.

This option is case insensitive and the default value is ACKNOWLEDGED.

hibernate.ogm.mongodb.write_concern_type

Specifies a custom WriteConcern implementation type (fully-qualified name, class object or instance). This is useful in cases where the pre-defined configurations are not sufficient, e.g. if you want to ensure that writes are propagated to a specific number of replicas or given "tag set". Only takes effect if hibernate.ogm.mongodb.write_concern is set to CUSTOM.

hibernate.ogm.mongodb.read_preference

Specifies the ReadPreference to be applied when issuing reads against the MongoDB datastore. Possible settings are (values of the ReadPreferenceType enum): PRIMARY, PRIMARY_PREFERRED, SECONDARY, SECONDARY_PREFERRED and NEAREST. It’s currently not possible to plug in custom read preference types. If you’re interested in such a feature, please let us know.

For more information, please refer to the official documentation.

When bootstrapping a session factory or entity manager factory programmatically, you should use the constants accessible via org.hibernate.ogm.datastore.mongodb.MongoDBProperties when specifying the configuration properties listed above.

Common properties shared between stores are declared on OgmProperties (a super interface of MongoDBProperties).

For maximum portability between stores, use the most generic interface possible.

10.2.3. Annotation based configuration

Hibernate OGM allows to configure store-specific options via Java annotations. You can override global configurations for a specific entity or even a specify property by virtue of the location where you place that annotation.

When working with the MongoDB backend, you can specify the following settings:

  • the write concern for entities and associations using the @WriteConcern annotation

  • the read preference for entities and associations using the @ReadPreference annotation

  • a strategy for storing associations using the @AssociationStorage and @AssociationDocumentStorage annotations

  • a strategy for storing the contents of map-typed associations using the @MapStorage annotation

Refer to <<mongodb-associations> to learn more about the options related to storing associations.

The following shows an example:

Example 49. Configuring the association storage strategy using annotations
@Entity
@WriteConcern(WriteConcernType.JOURNALED)
@ReadPreference(ReadPreferenceType.PRIMARY_PREFERRED)
@AssociationStorage(AssociationStorageType.ASSOCIATION_DOCUMENT)
@AssociationDocumentStorage(AssociationDocumentStorageType.COLLECTION_PER_ASSOCIATION)
@MapStorage(MapStorageType.AS_LIST)
public class Zoo {

    @OneToMany
    private Set<Animal> animals;

    @OneToMany
    private Set<Person> employees;

    @OneToMany
    @AssociationStorage(AssociationStorageType.IN_ENTITY)
    private Set<Person> visitors;

    // getters, setters ...
}

The @WriteConcern annotation on the entity level expresses that all writes should be done using the JOURNALED setting. Similarly, the @ReadPreference annotation advices the engine to preferably read that entity from the primary node if possible. The other two annotations on the type-level specify that all associations of the Zoo class should be stored in separate assocation documents, using a dedicated collection per association. This setting applies to the animals and employees associations. Only the elements of the visitors association will be stored in the document of the corresponding Zoo entity as per the configuration of that specific property which takes precedence over the entity-level configuration.

10.2.4. Programmatic configuration

In addition to the annotation mechanism, Hibernate OGM also provides a programmatic API for applying store-specific configuration options. This can be useful if you can’t modify certain entity types or don’t want to add store-specific configuration annotations to them. The API allows set options in a type-safe fashion on the global, entity and property levels.

When working with MongoDB, you can currently configure the following options using the API:

  • write concern

  • read preference

  • association storage strategy

  • association document storage strategy

  • strategy for storing the contents of map-typed associations

To set these options via the API, you need to create an OptionConfigurator implementation as shown in the following example:

Example 50. Example of an option configurator
public class MyOptionConfigurator extends OptionConfigurator {

    @Override
    public void configure(Configurable configurable) {
        configurable.configureOptionsFor( MongoDB.class )
            .writeConcern( WriteConcernType.REPLICA_ACKNOWLEDGED )
            .readPreference( ReadPreferenceType.NEAREST )
            .entity( Zoo.class )
                .associationStorage( AssociationStorageType.ASSOCIATION_DOCUMENT )
                .associationDocumentStorage( AssociationDocumentStorageType.COLLECTION_PER_ASSOCIATION )
                .mapStorage( MapStorageType.ASLIST )
                .property( "animals", ElementType.FIELD )
                    .associationStorage( AssociationStorageType.IN_ENTITY )
            .entity( Animal.class )
                .writeConcern( new RequiringReplicaCountOf( 3 ) )
                .associationStorage( AssociationStorageType.ASSOCIATION_DOCUMENT );
    }
}

The call to configureOptionsFor(), passing the store-specific identifier type MongoDB, provides the entry point into the API. Following the fluent API pattern, you then can configure global options (writeConcern(), readPreference()) and navigate to single entities or properties to apply options specific to these (associationStorage() etc.). The call to writeConcern() for the Animal entity shows how a specific write concern type can be used. Here RequiringReplicaCountOf is a custom implementation of WriteConcern which ensures that writes are propagated to a given number of replicas before a write is acknowledged.

Options given on the property level precede entity-level options. So e.g. the animals association of the Zoo class would be stored using the in entity strategy, while all other associations of the Zoo entity would be stored using separate association documents.

Similarly, entity-level options take precedence over options given on the global level. Global-level options specified via the API complement the settings given via configuration properties. In case a setting is given via a configuration property and the API at the same time, the latter takes precedence.

Note that for a given level (property, entity, global), an option set via annotations is overridden by the same option set programmatically. This allows you to change settings in a more flexible way if required.

To register an option configurator, specify its class name using the hibernate.ogm.option.configurator property. When bootstrapping a session factory or entity manager factory programmatically, you also can pass in an OptionConfigurator instance or the class object representing the configurator type.

10.3. Storage principles

Hibernate OGM tries to make the mapping to the underlying datastore as natural as possible so that third party applications not using Hibernate OGM can still read and update the same datastore. We worked particularly hard on the MongoDB model to offer various classic mappings between your object model and the MongoDB documents.

To describe things simply, each entity is stored as a MongoDB document. This document is stored in a MongoDB collection named after the entity type. The navigational information for each association from one entity to (a set of) entity is stored in the document representing the entity we are departing from.

10.3.1. Properties and built-in types

Each entity is represented by a document. Each property or more precisely column is represented by a field in this document, the field name being the column name.

Hibernate OGM supports by default the following property types:

  • java.lang.String

  { "text" : "Hello world!" }
  • java.lang.Character (or char primitive)

  { "delimiter" : "/" }
  • java.lang.Boolean (or boolean primitive)

  { "favorite" : true } # default mapping
  { "favorite" : "T" } # if @Type(type = "true_false") is given
  { "favorite" : "Y" } # if @Type(type = "yes_no") is given
  { "favorite" : 1 } # if @Type(type = "numeric_boolean") is given
  • java.lang.Byte (or byte primitive)

  { "display_mask" : "70" }
  • java.lang.Byte[] (or byte[])

  { "pdfAsBytes" : BinData(0,"MTIzNDU=") }
  • java.lang.Short (or short primitive)

  { "urlPort" : 80 }
  • java.lang.Integer (or integer primitive)

  { "stockCount" : 12309 }
  • java.lang.Long (or long primitive)

  { "userId" : NumberLong("-6718902786625749549") }
  • java.lang.Float (or float primitive)

  { "visitRatio" : 10.39 }
  • java.lang.Double (or double primitive)

  { "tax_percentage" : 12.34 }
  • java.math.BigDecimal

  { "site_weight" : "21.77" }
  • java.math.BigInteger

  { "site_weight" : "444" }
  • java.util.Calendar

  { "creation" : "2014/11/03 16:19:49:283 +0000" }
  • java.util.Date

  { "last_update" : ISODate("2014-11-03T16:19:49.283Z") }
  • java.util.UUID

  { "serialNumber" : "71f5713d-69c4-4b62-ad15-aed8ce8d10e0" }
  • java.util.URL

  { "url" : "http://www.hibernate.org/" }
  • org.bson.types.ObjectId

  { "object_id" : ObjectId("547d9b40e62048750f25ef77") }

Hibernate OGM doesn’t store null values in MongoDB, setting a value to null is the same as removing the field in the corresponding object in the db.

This can have consequences when it comes to queries on null value.

10.3.2. Entities

Entities are stored as MongoDB documents and not as BLOBs: each entity property will be translated into a document field. You can use @Table and @Column annotations to rename respectively the collection the document is stored in and the document’s field a property is persisted in.

Example 51. Default JPA mapping for an entity
@Entity
public class News {

    @Id
    private String id;
    private String title;

    // getters, setters ...
}
// Stored in the Collection "News"
{
    "_id" : "1234-5678-0123-4567",
    "title": "On the merits of NoSQL",
}
Example 52. Rename field and collection using @Table and @Column
@Entity
// Overrides the collection name
@Table(name = "News_Collection")
public class News {

    @Id
    private String id;

    // Overrides the field name
    @Column(name = "headline")
    private String title;

    // getters, setters ...
}
// Stored in the Collection "News"
{
    "_id" : "1234-5678-0123-4567",
    "headline": "On the merits of NoSQL",
}
Identifiers

Hibernate OGM always store identifiers using the _id field of a MongoDB document ignoring the name of the property in the entity.

That’s a good thing as MongoDB has special treatment and expectation of the property _id.

An identifier type may be one of the built-in types or a more complex type represented by an embedded class. When you use a built-in type, the identifier is mapped like a regular property. When you use an embedded class, then the _id is representing a nested document containing the embedded class properties.

Example 53. Define an identifier as a primitive type
@Entity
public class Bookmark {

    @Id
    private String id;

    private String title;

    // getters, setters ...
}
{
  "_id" : "bookmark_1"
  "title" : "Hibernate OGM documentation"
}
Example 54. Define an identifier using @EmbeddedId
@Embeddable
public class NewsID implements Serializable {

    private String title;
    private String author;

    // getters, setters ...
}

@Entity
public class News {

    @EmbeddedId
    private NewsID newsId;
    private String content;

    // getters, setters ...
}

News collection as JSON in MongoDB

{
  "_id" : {
      "author" : "Guillaume",
      "title" : "How to use Hibernate OGM ?"
  },
  "content" : "Simple, just like ORM but with a NoSQL database"
}

Generally, it is recommended though to work with MongoDB’s object id data type. This will facilitate the integration with other applications expecting that common MongoDB id type. To do so, you have two options:

  • Define your id property as org.bson.types.ObjectId

  • Define your id property as String and annotate it with @Type(type="objectid")

In both cases the id will be stored as native ObjectId in the datastore.

Example 55. Define an id as ObjectId
@Entity
public class News {

    @Id
    private ObjectId id;

    private String title;

    // getters, setters ...
}
Example 56. Define an id of type String as ObjectId
@Entity
public class News {

    @Id
    @Type(type = "objectid")
    private String id;

    private String title;

    // getters, setters ...
}
Identifier generation strategies

You can assign id values yourself or let Hibernate OGM generate the value using the @GeneratedValue annotation.

There are 4 different strategies:

  1. IDENTITY (suggested)

  2. TABLE

  3. SEQUENCE

  4. AUTO

1) IDENTITY generation strategy

The preferable strategy, Hibernate OGM will create the identifier upon insertion. To apply this strategy the id must be one of the following:

  • annotated with @Type(type="objectid")

  • org.bson.types.ObjectId

like in the following examples:

Example 57. Define an id of type String as ObjectId
@Entity
public class News {

    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    @Type(type = "objectid")
    private String id;

    private String title;

    // getters, setters ...
}
{
    "_id" : ObjectId("5425448830048b67064d40b1"),
    "title" : "Exciting News"
}
Example 58. Define an id as ObjectId
@Entity
public class News {

    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    private ObjectId id;

    private String title;

    // getters, setters ...
}
{
    "_id" : ObjectId("5425448830048b67064d40b1"),
    "title" : "Exciting News"
}

2) TABLE generation strategy

Example 59. Id generation strategy TABLE using default values
@Entity
public class GuitarPlayer {

    @Id
    @GeneratedValue(strategy = GenerationType.TABLE)
    private Long id;

    private String name;

    // getters, setters ...
}

GuitarPlayer collection

{
    "_id" : NumberLong(1),
    "name" : "Buck Cherry"
}

hibernate_sequences collection

{
    "_id" : "GuitarPlayer",
    "next_val" : 101
}
Example 60. Id generation strategy TABLE using a custom table
@Entity
public class GuitarPlayer {

    @Id
    @GeneratedValue(strategy = GenerationType.TABLE, generator = "guitarGen")
    @TableGenerator(
        name = "guitarGen",
        table = "GuitarPlayerSequence",
        pkColumnValue = "guitarPlayer",
        valueColumnName = "nextGuitarPlayerId"
    )
    private long id;

    // getters, setters ...
}

GuitarPlayer collection

{
    "_id" : NumberLong(1),
    "name" : "Buck Cherry"
}

GuitarPlayerSequence collection

{
    "_id" : "guitarPlayer",
    "nextGuitarPlayerId" : 2
}

3) SEQUENCE generation strategy

Example 61. SEQUENCE id generation strategy using default values
@Entity
public class Song {

  @Id
  @GeneratedValue(strategy = GenerationType.SEQUENCE)
  private Long id;

  private String title;

  // getters, setters ...
}

Song collection

{
  "_id" : NumberLong(2),
  "title" : "Flower Duet"
}

hibernate_sequences collection

{ "_id" : "song_sequence_name", "next_val" : 21 }
Example 62. SEQUENCE id generation strategy using custom values
@Entity
public class Song {

  @Id
  @GeneratedValue(strategy = GenerationType.SEQUENCE, generator = "songSequenceGenerator")
  @SequenceGenerator(
      name = "songSequenceGenerator",
      sequenceName = "song_seq",
      initialValue = 2,
      allocationSize = 20
  )
  private Long id;

  private String title;

  // getters, setters ...
}

Song collection

{
  "_id" : NumberLong(2),
  "title" : "Flower Duet"
}

hibernate_sequences collection

{ "_id" : "song_seq", "next_val" : 42 }

4) AUTO generation strategy

Care must be taken when using the GenerationType.AUTO strategy. When the property hibernate.id.new_generator_mappings is set to false (default), it will map to the IDENTITY strategy. As described before, this requires your ids to be of type ObjectId or @Type(type = "objectid") String. If hibernate.id.new_generator_mappings is set to true, AUTO will be mapped to the TABLE strategy. This requires your id to be of a numeric type.

We recommend to not use AUTO but one of the explicit strategies (IDENTITY or TABLE) to avoid potential misconfigurations.

For more details you can check the issue OGM-663.

If the property hibernate.id.new_generator_mappings is set to false, AUTO will behave as the IDENTITY strategy.

If the property hibernate.id.new_generator_mappings is set to true, AUTO will behave as the SEQUENCE strategy.

Example 63. AUTO id generation strategy using default values
@Entity
public class DistributedRevisionControl {

  @Id
  @GeneratedValue(strategy = GenerationType.AUTO)
  private Long id;

  private String name;

  // getters, setters ...
}

DistributedRevisionControl collection

{ "_id" : NumberLong(1), "name" : "Git" }

hibernate_sequences collection

{ "_id" : "hibernate_sequence", "next_val" : 2 }
Example 64. AUTO id generation strategy wih hibernate.id.new_generator_mappings set to false and ObjectId
@Entity
public class Comedian {

  @Id
  @GeneratedValue(strategy = GenerationType.AUTO)
  private ObjectId id;

  private String name;

  // getters, setters ...
}

Comedian collection

{ "_id" : ObjectId("5458b11693f4add0f90519c5"), "name" : "Louis C.K." }
Example 65. Entity with @EmbeddedId
@Entity
public class News {

    @EmbeddedId
    private NewsID newsId;

    // getters, setters ...
}

@Embeddable
public class NewsID implements Serializable {

    private String title;
    private String author;

    // getters, setters ...
}

Rendered as JSON in MongoDB

{
    "_id" :{
        "title": "How does Hibernate OGM MongoDB work?",
        "author": "Guillaume"
    }
}
Embedded objects and collections

Hibernate OGM stores elements annotated with @Embedded or @ElementCollection as nested documents of the owning entity.

Example 66. Embedded object
@Entity
public class News {

    @Id
    private String id;
    private String title;

    @Embedded
    private NewsPaper paper;

    // getters, setters ...
}

@Embeddable
public class NewsPaper {

    private String name;
    private String owner;

    // getters, setters ...
}
{
    "_id" : "1234-5678-0123-4567",
    "title": "On the merits of NoSQL",
    "paper": {
        "name": "NoSQL journal of prophecies",
        "owner": "Delphy"
    }
}
Example 67. @ElementCollection with primitive types
@Entity
public class AccountWithPhone {

    @Id
    private String id;

    @ElementCollection
    private List<String> mobileNumbers;

    // getters, setters ...
}

AccountWithPhone collection

{
    "_id" : "john_account",
    "mobileNumbers" : [ "+1-222-555-0222", "+1-202-555-0333" ]
}
Example 68. @ElementCollection with one attribute
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
{
    "_id" : "df153180-c6b3-4a4c-a7da-d5de47cf6f00",
    "grandChildren" : [ "Luke", "Leia" ]
}

The class GrandChild has only one attribute name, this means that Hibernate OGM doesn’t need to store the name of the attribute.

If the nested document has two or more fields, like in the following example, Hibernate OGM will store the name of the fields as well.

Example 69. @ElementCollection with @OrderColumn
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    @OrderColumn( name = "birth_order" )
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
{
    "_id" : "e3e1ed4e-c685-4c3f-9a67-a5aeec6ff3ba",
    "grandChildren" :
        [
            {
                "name" : "Luke",
                "birth_order" : 0
            },
            {
                "name" : "Leia",
                "birthorder" : 1
            }
        ]
}
Example 70. @ElementCollection with Map of @Embeddable
@Entity
public class ForumUser {

    @Id
    private String name;

    @ElementCollection
    private Map<String, JiraIssue> issues = new HashMap<>();

    // getters, setters ...
}

@Embeddable
public class JiraIssue {

    private Integer number;
    private String project;

    // getters, setters ...
}
{
        "_id" : "Jane Doe",
        "issues" : {
                "issueWithNull" : {
                },
                "issue2" : {
                    "number" : 2000,
                    "project" : "OGM"
                },
                "issue1" : {
                    "number" : 1253,
                    "project" : "HSEARCH"
                }
        }
}

You can override the column name used for a property of an embedded object. But you need to know that the default column name is the concatenation of the embedding property, a . (dot) and the embedded property (recursively for several levels of embedded objects).

The MongoDB datastore treats dots specifically as it transforms them into nested documents. If you want to override one column name and still keep the nested structure, don’t forget the dots.

That’s a bit abstract, so let’s use an example.

@Entity
class Order {
    @Id String number;
    User user;
    Address shipping;
    @AttributeOverrides({
        @AttributeOverride(name="name", column=@Column(name="delivery.provider"),
        @AttributeOverride(name="expectedDelaysInDays", column=@Column(name="delivery.delays")
    })
    DeliveryProvider deliveryProvider;
    CreditCardType cardType;
}

// default columns
@Embedded
class User {
    String firstname;
    String lastname;
}

// override one column
@Embeddable
public Address {
    String street;
    @Column(name="shipping.dest_city")
    String city;
}

// both columns overridden from the embedding side
@Embeddable
public DeliveryProvider {
    String name;
    Integer expectedDelaysInDays;
}

// do not use dots in the overriding
// and mix levels (bad form)
@Embedded
class CreditCardType {
    String merchant;
    @Column(name="network")
    String network;
}
{
    "_id": "123RF33",
    "user": {
        "firstname": "Emmanuel",
        "lastname": "Bernard"
    },
    "shipping": {
        "street": "1 av des Champs Elysées",
        "dest_city": "Paris"
    },
    "delivery": {
        "provider": "Santa Claus Inc.",
        "delays": "1"
    }
    "network": "VISA",
    "cardType: {
        "merchant": "Amazon"
    }
}

If you share the same embeddable in different places, you can use JPA’s @AttributeOverride to override columns from the embedding side. This is the case of DeliveryProvider in our example.

If you omit the dot in one of the columns, this column will not be part of the nested document. This is demonstrated by the CreditCardType. We advise you against it. Like crossing streams, it is bad form. This approach might not be supported in the future.

10.3.3. Associations

Hibernate OGM MongoDB proposes three strategies to store navigation information for associations. The three possible strategies are:

To switch between these strategies, use of the three approaches to options:

  • annotate your entity with @AssocationStorage and @AssociationDocumentStorage annotations (see Annotation based configuration),

  • use the API for programmatic configuration (see Programmatic configuration)

  • or specify a default strategy via the hibernate.ogm.datastore.document.association_storage and hibernate.ogm.mongodb.association_document_storage configuration properties.

In Entity strategy

In this strategy, Hibernate OGM stores the id(s) of the associated entity(ies) into the entity document itself. This field stores the id value for to-one associations and an array of id values for to-many associations. An embedded id will be represented by a nested document. For indexed collections (i.e. List or Map), the index will be stored along the id.

When using this strategy the annotations @JoinTable will be ignored because no collection is created for associations.

You can use @JoinColumn to change the name of the field that stores the foreign key (as an example, see Unidirectional one-to-one with @JoinColumn).

To-one associations
Example 71. Unidirectional one-to-one
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}


@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    private Vehicule vehicule;

    // getters, setters ...
}
{
  "_id" : "V_01",
  "brand" : "Mercedes"
}

Wheel collection as JSON in MongoDB

{
  "_id" : "W001",
  "diameter" : 0,
  "vehicule_id" : "V_01"
}
Example 72. Unidirectional one-to-one with @JoinColumn
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}


@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    @JoinColumn( name = "part_of" )
    private Vehicule vehicule;

    // getters, setters ...
}
{
  "_id" : "V_01",
  "brand" : "Mercedes"
}

Wheel collection as JSON in MongoDB

{
  "_id" : "W001",
  "diameter" : 0,
  "part_of" : "V_01"
}

In a true one-to-one association, it is possible to share the same id between the two entities and therefore a foreign key is not required. You can see how to map this type of association in the following example:

Example 73. Unidirectional one-to-one with @MapsId and @PrimaryKeyJoinColumn
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}

@Entity
public class Wheel {

    @Id
    private String id;
    private double diameter;

    @OneToOne
    @PrimaryKeyJoinColumn
    @MapsId
    private Vehicule vehicule;

    // getters, setters ...
}

Vehicule collection as JSON in MongoDB

{
  "_id" : "V_01",
  "brand" : "Mercedes"
}

Wheel collection as JSON in MongoDB

{
  "_id" : "V_01",
  "diameter" : 0,
}
Example 74. Bidirectional one-to-one
@Entity
public class Husband {

    @Id
    private String id;
    private String name;

    @OneToOne
    private Wife wife;

    // getters, setters ...
}

@Entity
public class Wife {

    @Id
    private String id;
    private String name;

    @OneToOne
    private Husband husband;

    // getters, setters ...
}

Husband collection as JSON in MongoDB

{
  "_id" : "alex",
  "name" : "Alex",
  "wife" : "bea"
}

Wife collection as JSON in MongoDB

{
  "_id" : "bea",
  "name" : "Bea",
  "husband" : "alex"
}
Example 75. Unidirectional many-to-one
@Entity
public class JavaUserGroup {

    @Id
    private String jugId;
    private String name;

    // getters, setters ...
}

@Entity
public class Member {

    @Id
    private String id;
    private String name;

    @ManyToOne
    private JavaUserGroup memberOf;

    // getters, setters ...
}

JavaUserGroup collection as JSON in MongoDB

{
    "_id" : "summer_camp",
    "name" : "JUG Summer Camp"
}

Member collection as JSON in MongoDB

{
    "_id" : "jerome",
    "name" : "Jerome"
    "memberOf_jugId" : "summer_camp"
}
{
    "_id" : "emmanuel",
    "name" : "Emmanuel Bernard"
    "memberOf_jugId" : "summer_camp"
}
Example 76. Bidirectional many-to-one
@Entity
public class SalesForce {

    @Id
    private String id;
    private String corporation;

    @OneToMany(mappedBy = "salesForce")
    private Set<SalesGuy> salesGuys = new HashSet<SalesGuy>();

    // getters, setters ...
}

@Entity
public class SalesGuy {
    private String id;
    private String name;

    @ManyToOne
    private SalesForce salesForce;

    // getters, setters ...
}

SalesForce collection

{
    "_id" : "red_hat",
    "corporation" : "Red Hat",
    "salesGuys" : [ "eric", "simon" ]
}

SalesGuy collection

{
    "_id" : "eric",
    "name" : "Eric"
    "salesForce_id" : "red_hat",
}
{
    "_id" : "simon",
    "name" : "Simon",
    "salesForce_id" : "red_hat"
}
Example 77. Bidirectional many-to-one between entities with embedded ids
@Entity
public class Game {

    @EmbeddedId
    private GameId id;

    private String name;

    @ManyToOne
    private Court playedOn;

    // getters, setters ...
}


public class GameId implements Serializable {

    private String category;

    @Column(name = "id.gameSequenceNo")
    private int sequenceNo;

    // getters, setters ...
    // equals / hashCode
}

@Entity
public class Court {

    @EmbeddedId
    private CourtId id;

    private String name;

    @OneToMany(mappedBy = "playedOn")
    private Set<Game> games = new HashSet<Game>();

    // getters, setters ...
}

public class CourtId implements Serializable {

    private String countryCode;
    private int sequenceNo;

    // getters, setters ...
    // equals / hashCode
}
Court collection
{
    "_id" : {
        "countryCode" : "DE",
        "sequenceNo" : 123
    },
    "name" : "Hamburg Court",
    "games" : [
        { "gameSequenceNo" : 457, "category" : "primary" },
        { "gameSequenceNo" : 456, "category" : "primary" }
    ]
}
Game collection
{
    "_id" : {
        "category" : "primary",
        "gameSequenceNo" : 456
    },
    "name" : "The game",
    "playedOn_id" : {
        "countryCode" : "DE",
        "sequenceNo" : 123
    }
}
{
    "_id" : {
        "category" : "primary",
        "gameSequenceNo" : 457
    },
    "name" : "The other game",
    "playedOn_id" : {
        "countryCode" : "DE",
        "sequenceNo" : 123
    }
}

Here we see that the embedded id is represented as a nested document and directly referenced by the associations.

To-many associations
Example 78. Unidirectional one-to-many
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide",
  "products" : [ "Beer", "Pretzel" ]
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}
Example 79. Unidirectional one-to-many with @OrderColumn
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide",
  "products" : [
    {
      "products_name" : "Pretzel",
      "products_ORDER" : 1
    },
    {
      "products_name" : "Beer",
      "products_ORDER" : 0
    }
  ]
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}

A map can be used to represent an association, in this case Hibernate OGM will store the key of the map and the associated id.

Example 80. Unidirectional one-to-many using maps with defaults
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    private Map<String, Address> addresses = new HashMap<String, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}

User collection as JSON in MongoDB

{
  "_id" : "user_001",
  "addresses" : [
    {
      "work" : "address_001",
      "home" : "address_002"
    }
  ]
}

Address collection as JSON in MongoDB

{ "_id" : "address_001", "city" : "Rome" }
{ "_id" : "address_002", "city" : "Paris" }

If the map value cannot be represented by a single field (e.g. when referencing a type with a composite id or using an embeddable type as map value type), a sub-document containing all the required fields will be stored as value.

If the map key either is not of type String or it is made up of several columns (composite map key), the optimized structure shown in the example above cannot be used as MongoDB only allows for Strings as field names. In that case the association will be represented by a list of sub-documents, also containing the map key column(s). You can use @MapKeyColumn to rename the field containing the key of the map, otherwise it will default to "<%COLLECTION_ROLE%>_KEY", e.g. "addresses_KEY".

Example 81. Unidirectional one-to-many using maps with @MapKeyColumn
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    @MapKeyColumn(name = "addressType")
    private Map<Long, Address> addresses = new HashMap<Long, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}

User collection as JSON in MongoDB

{
  "_id" : "user_001",
  "addresses" : [
    {
      "addressType" : 1,
      "addresses_id" : "address_001"
    },
    {
      "addressType" : 2,
      "addresses_id" : "address_002"
    }
  ]
}

Address collection as JSON in MongoDB

{ "_id" : "address_001", "city" : "Rome" }
{ "_id" : "address_002", "city" : "Paris" }

In case you want to enforce the list-style represention also for maps with a single key column of type String (e.g. when reading back data persisted by earlier versions of Hibernate OGM), you can do so by setting the option hibernate.ogm.datastore.document.map_storage to the value AS_LIST.

Example 82. Unidirectional many-to-many using in entity strategy
@Entity
public class Student {

    @Id
    private String id;
    private String name;

    // getters, setters ...
}

@Entity
public class ClassRoom {

    @Id
    private long id;
    private String lesson;

    @ManyToMany
    private List<Student> students = new ArrayList<Student>();

    // getters, setters ...
}

Student collection

{
  "_id" : "john",
  "name" :"John Doe" }
{
  "_id" : "mario",
  "name" : "Mario Rossi"
}
{
  "_id" : "kate",
  "name" : "Kate Doe"
}

ClassRoom collection

{
  "_id" : NumberLong(1),
  "lesson" : "Math"
  "students" : [
     "mario",
     "john"
  ]
}
{
  "_id" : NumberLong(2),
  "lesson" : "English"
  "students" : [
     "mario",
     "kate"
  ]
}
Example 83. Bidirectional many-to-many
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany( mappedBy = "bankAccounts" )
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}

AccountOwner collection

{
    "_id" : "owner_1",
    "SSN" : "0123456"
    "bankAccounts" : [ "account_1" ]
}

BankAccount collection

{
    "_id" : "account_1",
    "accountNumber" : "X2345000"
    "owners" : [ "owner_1", "owner2222" ]
}
Example 84. Ordered list with embedded id
@Entity
public class Race {
    @EmbeddedId
    private RaceId raceId;

    @OrderColumn(name = "ranking")
    @OneToMany @JoinTable(name = "Race_Runners")
    private List<Runner> runnersByArrival = new ArrayList<Runner>();

    // getters, setters ...
}

public class RaceId implements Serializable {
    private int federationSequence;
    private int federationDepartment;

    // getters, setters, equals, hashCode
}

@Entity
public class Runner {
    @EmbeddedId
    private RunnerId runnerId;
    private int age;

    // getters, setters ...
}

public class RunnerId implements Serializable {
    private String firstname;
    private String lastname;

    // getters, setters, equals, hashCode
}
Race collection
{
    "_id": {
        "federationDepartment": 75,
        "federationSequence": 23
    },
    "runnersByArrival": [{
        "firstname": "Pere",
        "lastname": "Noel",
        "ranking": 1
    }, {
        "firstname": "Emmanuel",
        "lastname": "Bernard",
        "ranking": 0
    }]
}
Runner collection
{
    "_id": {
        "firstname": "Pere",
        "lastname": "Noel"
    },
    "age": 105
} {
    "_id": {
        "firstname": "Emmanuel",
        "lastname": "Bernard"
    },
    "age": 37
}
One collection per association strategy

In this strategy, Hibernate OGM creates a MongoDB collection per association in which it will store all navigation information for that particular association.

This is the strategy closest to the relational model. If an entity A is related to B and C, 2 collections will be created. The name of this collection is made of the association table concatenated with associations_.

For example, if the BankAccount and Owner are related, the collection used to store will be named associations_Owner_BankAccount. You can rename The prefix is useful to quickly identify the association collections from the entity collections. You can also decide to rename the collection representing the association using @JoinTable (see an example)

Each document of an association collection has the following structure:

  • _id contains the id of the owner of relationship

  • rows contains all the id of the related entities

The preferred approach is to use the in-entity strategy but this approach can alleviate the problem of having documents that are too big.

Example 85. Unidirectional relationship
{
    "_id" : { "owners_id" : "owner0001" },
    "rows" : [
        "accountABC",
        "accountXYZ"
    ]
}
Example 86. Bidirectional relationship
{
    "_id" : { "owners_id" : "owner0001" },
    "rows" : [ "accountABC", "accountXYZ" ]
}
{
    "_id" : { "bankAccounts_id" : "accountXYZ" },
    "rows" : [ "owner0001" ]
}

This strategy won’t affect *-to-one associations or embedded collections.

Example 87. Unidirectional one-to-many using one collection per strategy
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide"
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}

associations_Basket_Product collection

{
  "_id" : { "Basket_id" : "davide_basket" },
  "rows" : [ "Beer", "Pretzel" ]
}

The order of the element in the list might be preserved using @OrderColumn. Hibernate OGM will store the order adding an additional fieldd to the document containing the association.

Example 88. Unidirectional one-to-many using one collection per strategy with @OrderColumn
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    @OrderColumn
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide"
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}

associations_Basket_Product collection

{
  "_id" : { "Basket_id" : "davide_basket" },
  "rows" : [
    {
      "products_name" : "Pretzel",
      "products_ORDER" : 1
    },
    {
      "products_name" : "Beer",
      "products_ORDER" : 0
    }
  ]
}
Example 89. Unidirectional many-to-many using one collection per association strategy
@Entity
public class Student {

    @Id
    private String id;
    private String name;

    // getters, setters ...
}

@Entity
public class ClassRoom {

    @Id
    private long id;
    private String lesson;

    @ManyToMany
    private List<Student> students = new ArrayList<Student>();

    // getters, setters ...
}

Student collection

{
  "_id" : "john",
  "name" : "John Doe"
}
{
  "_id" : "mario",
  "name" : "Mario Rossi"
}
{
  "_id" : "kate",
  "name" : "Kate Doe"
}

ClassRoom collection

{
  "_id" : NumberLong(1),
  "lesson" : "Math"
}
{
  "_id" : NumberLong(2),
  "lesson" : "English"
}

associations_ClassRoom_Student

{
  "_id" : {
    "ClassRoom_id" : NumberLong(1),
  },
  "rows" : [ "john", "mario" ]
}
{
  "_id" : {
    "ClassRoom_id" : NumberLong(2),
  },
  "rows" : [ "mario", "kate" ]
}
Example 90. Bidirectional many-to-many using one collection per association strategy
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany(mappedBy = "bankAccounts")
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}

AccountOwner collection

{
  "_id" : "owner_1",
  "SSN" : "0123456"
}

BankAccount collection

{
  "_id" : "account_1",
  "accountNumber" : "X2345000"
}

associations_AccountOwner_BankAccount collection

{
  "_id" : {
    "bankAccounts_id" : "account_1"
  },
  "rows" : [ "owner_1" ]
}
{
  "_id" : {
    "owners_id" : "owner_1"
  },
  "rows" : [ "account_1" ]
}

You can change the name of the collection containing the association using the @JoinTable annotation. In the following example, the name of the collection containing the association is OwnerBankAccounts (instead of the default associations_AccountOwner_BankAccount)

Example 91. Bidirectional many-to-many using one collection per association strategy and @JoinTable
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    @JoinTable( name = "OwnerBankAccounts" )
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany(mappedBy = "bankAccounts")
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}

AccountOwner collection

{
  "_id" : "owner_1",
  "SSN" : "0123456"
}

BankAccount collection

{
  "_id" : "account_1",
  "accountNumber" : "X2345000"
}

OwnerBankAccount

{
  "_id" : {
    "bankAccounts_id" : "account_1"
  },
  "rows" : [ "owner_1" ]
}
{
  "_id" : {
    "owners_id" : "owner_1"
  },
  "rows" : [ "account_1" ]
}
Global collection strategy

With this strategy, Hibernate OGM creates a single collection named Associations in which it will store all navigation information for all associations. Each document of this collection is structured in 2 parts. The first is the _id field which contains the identifier information of the association owner and the name of the association table. The second part is the rows field which stores (into an embedded collection) all ids that the current instance is related to.

This strategy won’t affect *-to-one associations or embedded collections.

Generally, you should not make use of this strategy unless embedding the association information proves to be too big for your document and you wish to separate them.

Example 92. Associations collection containing unidirectional association
{
    "_id": {
        "owners_id": "owner0001",
        "table": "AccountOwner_BankAccount"
    },
    "rows": [ "accountABC", "accountXYZ" ]
}

For a bidirectional relationship, another document is created where ids are reversed. Don’t worry, Hibernate OGM takes care of keeping them in sync:

Example 93. Associations collection containing a bidirectional association
{
    "_id": {
        "owners_id": "owner0001",
        "table": "AccountOwner_BankAccount"
    },
    "rows": [ "accountABC", "accountXYZ" ]
}
{
    "_id": {
        "bankAccounts_id": "accountXYZ",
        "table": "AccountOwner_BankAccount"
    },
    "rows": [ "owner0001" ]
}
Example 94. Unidirectional one-to-many using global collection strategy
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide"
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}

Associations collection

{
  "_id" : {
    "Basket_id" : "davide_basket",
    "table" : "Basket_Product"
  },
  "rows" : [
    {
      "products_name" : "Pretzel",
      "products_ORDER" : 1
    },
    {
      "products_name" : "Beer",
    "products_ORDER" : 0
    }
  ]
}
Example 95. Unidirectional one-to-many using global collection strategy with @JoinTable
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    // It will change the value stored in the field table in the Associations collection
    @JoinTable( name = "BasketContent" )
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}

Basket collection

{
  "_id" : "davide_basket",
  "owner" : "Davide"
}

Product collection

{
  "_id" : "Pretzel",
  "description" : "Glutino Pretzel Sticks"
}
{
  "_id" : "Beer",
  "description" : "Tactical nuclear penguin"
}

Associations collection

{
  "_id" : {
    "Basket_id" : "davide_basket",
    "table" : "BasketContent"
  },
  "rows" : [ "Beer", "Pretzel" ]
}
Example 96. Unidirectional many-to-many using global collection strategy
@Entity
public class Student {

    @Id
    private String id;
    private String name;

    // getters, setters ...
}

@Entity
public class ClassRoom {

    @Id
    private long id;
    private String lesson;

    @ManyToMany
    private List<Student> students = new ArrayList<Student>();

    // getters, setters ...
}

Student collection

{
  "_id" : "john",
  "name" : "John Doe"
}
{
  "_id" : "mario",
  "name" : "Mario Rossi"
}
{
  "_id" : "kate",
  "name" : "Kate Doe"
}

ClassRoom collection

{
  "_id" : NumberLong(1),
  "lesson" : "Math"
}
{
  "_id" : NumberLong(2),
  "lesson" : "English"
}

Associations collection

{
  "_id" : {
    "ClassRoom_id" : NumberLong(1),
    "table" : "ClassRoom_Student"
  },
  "rows" : [ "john", "mario" ]
}
{
  "_id" : {
    "ClassRoom_id" : NumberLong(2),
    "table" : "ClassRoom_Student"
  },
  "rows" : [ "mario", "kate" ]
}
Example 97. Bidirectional many-to-many using global collection strategy
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany(mappedBy = "bankAccounts")
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}

AccountOwner collection

{
  "_id" : "owner0001",
  "SSN" : "0123456"
}

BankAccount collection

{
  "_id" : "account_1",
  "accountNumber" : "X2345000"
}

Associations collection

{
  "_id" : {
    "bankAccounts_id" : "account_1",
    "table" : "AccountOwner_BankAccount"
    },

  "rows" : [ "owner0001" ]
}
{
  "_id" : {
    "owners_id" : "owner0001",
    "table" : "AccountOwner_BankAccount"
  },

  "rows" : [ "account_1" ]
}

10.4. Indexes and unique constraints

10.4.1. Standard indexes and unique constraints

You can create your index and unique constraints in MongoDB using the standard JPA annotations.

Example 98. Creating indexes and unique constraints using JPA annotations
@Entity
@Table(indexes = {
    @Index(columnList = "author, name", name = "author_name_idx", unique = true),
    @Index(columnList = "name DESC", name = "name_desc_idx")
})
public class Poem {

    @Id
    private String id;
    private String name;
    private String author;

    @Column(unique = true)
    private String url;

   // getters, setters ...
}

MongoDB supports unique constraints via unique indexes. It considers null as a value to be unique: you can only have one null value per unique index. This is not what is commonly accepted as the definition of a unique constraint in the JPA world. Thus, by default, we create the unique indexes as sparse: it only indexes defined values so that the unique constraints accept multiple null values.

10.4.2. Using MongoDB specific index options

It is possible to define them using the @IndexOption annotation.

Example 99. Creating indexes with MongoDB specific options
@Entity
@Table(indexes = {
    @Index(columnList = "author", name = "author_idx")
})
@IndexOptions(
    @IndexOption(forIndex = "author_idx", options = "{ background : true, sparse : true, partialFilterExpression : { author: 'Verlaine' } }")
)
public class Poem {

    @Id
    private String id;
    private String name;
    private String author;

   // getters, setters ...
}

@IndexOption simply passes the options to MongoDB at index creation: you can use every option available in MongoDB.

10.4.3. Full text indexes

MongoDB supports the ability to create one (and only one) full text index per collection.

As JPA does not support the ability to define text as an order in the @Index annotation (only ASC and DESC are supported), this ability has been included inside the @IndexOption mechanism. You simply need to add text: true to the options passed to MongoDB, Hibernate OGM interprets it and translates the index to a full text index.

Example 100. Creating a full text index
@Entity
@Table(indexes = {
    @Index(columnList = "author, name", name = "author_name_text_idx")
})
@IndexOptions(
    @IndexOption(forIndex = "author_name_text_idx", options = "{ text: true, default_language : 'fr', weights : { author: 2, name: 5 } }")
)
public class Poem {

    @Id
    private String id;
    private String name;
    private String author;

   // getters, setters ...
}

10.5. Transactions

MongoDB does not support transactions. Only changes applied to the same document are done atomically. A change applied to more than one document will not be applied atomically. This problem is slightly mitigated by the fact that Hibernate OGM queues all changes before applying them during flush time. So the window of time used to write to MongoDB is smaller than what you would have done manually.

We recommend that you still use transaction demarcations with Hibernate OGM to trigger the flush operation transparently (on commit). But do not consider rollback as a possibility, this won’t work.

10.6. Optimistic Locking

MongoDB does not provide a built-in mechanism for detecting concurrent updates to the same document but it provides a way to execute atomic find and update operations. By exploiting this commands Hibernate OGM can detect concurrent modifications to the same document.

You can enable optimistic locking detection using the annotation @Version:

Example 101. Optimistic locking detection via @Version
@Entity
public class Planet implements Nameable {

    @Id
    private String id;
    private String name;

    @Version
    private int version;

   // getters, setters ...
}
{
  "_id" : "planet-1",
  "name" : "Pluto",
  "version" : 0
}

The @Version annotation define which attribute will keep track of the version of the document, Hibernate OGM will update the field when required and if two changes from two different sessions (for example) are applied to the same document a org.hibernate.StaleObjectStateException is thrown.

You can use @Column to change the name of the field created on MongoDB:

Example 102. Optimistic locking detection via @Version using @Column
@Entity
public class Planet implements Nameable {

    @Id
    private String id;
    private String name;

    @Version
    @Column(name="OPTLOCK")
    private int version;

   // getters, setters ...
}
{
  "_id" : "planet-1",
  "name" : "Pluto",
  "OPTLOCK" : 0
}

10.7. Queries

You can express queries in a few different ways:

  • using JPQL

  • using a native MongoQL query

  • using a Hibernate Search query (brings advanced full-text and geospatial queries)

While you can use JPQL for simple queries, you might hit limitations. The current recommended approach is to use native MongoQL if your query involves nested (list of) elements.

MongoDB doesn’t require Hibernate Search to run queries.

In order to reflect changes performed in the current session, all entities affected by a given query are flushed to the datastore prior to query execution (that’s the case for Hibernate ORM as well as Hibernate OGM).

For not fully transactional stores such as MongoDB this can cause changes to be written as a side-effect of running queries which cannot be reverted by a possible later rollback.

Depending on your specific use cases and requirements you may prefer to disable auto-flushing, e.g. by invoking query.setFlushMode( FlushMode.MANUAL ). Bear in mind though that query results will then not reflect changes applied within the current session.

10.7.1. JPQL queries

Hibernate OGM is a work in progress, so only a sub-set of JPQL constructs is available when using the JPQL query support. This includes:

  • simple comparisons using "<", "<=", "=", ">=" and ">"

  • IS NULL and IS NOT NULL

  • the boolean operators AND, OR, NOT

  • LIKE, IN and BETWEEN

  • ORDER BY

  • inner JOIN on embedded collections

  • projections of regular and embedded properties

Queries using these constructs will be transformed into equivalent native MongoDB queries.

Let us know by opening an issue or sending an email what query you wish to execute. Expanding our support in this area is high on our priority list.

10.7.2. Native MongoDB queries

Hibernate OGM also supports certain forms of native queries for MongoDB. Currently two forms of native queries are available via the MongoDB backend:

  • find queries specifying the search criteria only

  • queries specified using the MongoDB CLI syntax (CLI Syntax)

The former always maps results to entity types. The latter either maps results to entity types or to certain supported forms of projection. Note that parameterized queries are not supported by MongoDB, so don’t expect Query#setParameter() to work.

You can execute native queries as shown in the following example:

Example 103. Using the JPA API
@Entity
public class Poem {

    @Id
    private Long id;

    private String name;

    private String author;

   // getters, setters ...
}

...

javax.persistence.EntityManager em = ...

// criteria-only find syntax
String query1 = "{ $and: [ { name : 'Portia' }, { author : 'Oscar Wilde' } ] }";
Poem poem = (Poem) em.createNativeQuery( query1, Poem.class ).getSingleResult();

// criteria-only find syntax with order-by
String query2 = "{ $query : { author : 'Oscar Wilde' }, $orderby : { name : 1 } }";
List<Poem> poems = em.createNativeQuery( query2, Poem.class ).getResultList();

// projection via CLI-syntax
String query3 = "db.WILDE_POEM.find(" +
    "{ '$query' : { 'name' : 'Athanasia' }, '$orderby' : { 'name' : 1 } }" +
    "{ 'name' : 1 }" +
    ")";

// will contain name and id as MongoDB always returns the id for projections
List<Object[]> poemNames = (List<Object[]>)em.createNativeQuery( query3 ).getResultList();

// projection via CLI-syntax
String query4 = "db.WILDE_POEM.count({ 'name' : 'Athanasia' })";

Object[] count = (Object[])em.createNativeQuery( query4 ).getSingleResult();

The result of a query is a managed entity (or a list thereof) or a projection of attributes in form of an object array, just like you would get from a JPQL query.

Example 104. Using the Hibernate native API
OgmSession session = ...

String query1 = "{ $and: [ { name : 'Portia' }, { author : 'Oscar Wilde' } ] }";
Poem poem = session.createNativeQuery( query1 )
                      .addEntity( "Poem", Poem.class )
                      .uniqueResult();

String query2 = "{ $query : { author : 'Oscar Wilde' }, $orderby : { name : 1 } }";
List<Poem> poems = session.createNativeQuery( query2 )
                      .addEntity( "Poem", Poem.class )
                      .list();

Native queries can also be created using the @NamedNativeQuery annotation:

Example 105. Using @NamedNativeQuery
@Entity
@NamedNativeQuery(
   name = "AthanasiaPoem",
   query = "{ $and: [ { name : 'Athanasia' }, { author : 'Oscar Wilde' } ] }",
   resultClass = Poem.class )
public class Poem { ... }

...

// Using the EntityManager
Poem poem1 = (Poem) em.createNamedQuery( "AthanasiaPoem" )
                     .getSingleResult();

// Using the Session
Poem poem2 = (Poem) session.getNamedQuery( "AthanasiaPoem" )
                     .uniqueResult();

Hibernate OGM stores data in a natural way so you can still execute queries using the MongoDB driver, the main drawback is that the results are going to be raw MongoDB documents and not managed entities.

CLI Syntax

Specifying native MongoDB queries using the CLI syntax is an EXPERIMENTAL feature for the time being.

Hibernate OGM can execute native queries expressed using the MongoDB CLI syntax with some limitations. Currently find(), findOne(), findAndModify(), and count() queries are supported. Furthermore, three types of write queries are supported via the CLI syntax: insert(), remove(), and update(). Other query types may be supported in future versions.

As one would expect, find(), findOne(), findAndModify(), aggregate, distinct(), and count() can be executed using javax.persistence.Query.getSingleResult() or javax.persistence.Query.getResultList(), while insert(), remove(), and update() require using javax.persistence.Query.executeUpdate(). Also note that, javax.persistence.Query.executeUpdate() may return -1 in case execution of a query was not acknowledged relative to the write concern used.

The following functions can be used in the provided JSON: BinData, Date, HexData, ISODate, NumberLong, ObjectId, Timestamp, RegExp, DBPointer, UUID, GUID, CSUUID, CSGUID, JUUID, JGUID, PYUUID, PYGUID.

NumberInt is not supported as it is currently not supported by the MongoDB Java driver.

No cursor operations such as sort() are supported. Instead use the corresponding MongoDB query modifiers such as $orderby within the criteria parameter.

You can limit the results of a query using the setMaxResults(…​) method.

JSON parameters passed via the CLI syntax must be specified using the strict mode. Specifically, keys need to be given within quotes; the only relaxation of this is that single quotes may be used when specifying attribute names/values to facilitate embedding queries within Java strings.

Note that results of projections are returned as retrieved from the MongoDB driver at the moment and are not (yet) converted using suitable Hibernate OGM type implementations. This requirement is tracked under OGM-1031.

Example 106. CLI syntax examples
// Valid syntax
String valid = "db.Poem.find({ \"name\" : \"Athanasia\" })";

String alsoValid = "db.Poem.find({ '$or' : [{'name': 'Athanasia' }, {'name': 'Portia' }]})";

String validAggregation = "db.Poem.aggregate([{ '$match': {'author': { '$regex': 'oscar.*', '$options': 'i' } } }, { '$sort': {'name': -1 } } ])";

// NOT Valid syntax, it will throw an exception: com.mongodb.util.JSONParseException
String notValid =  "db.Poem.find({ name : \"Athanasia\" })";

String alsoNotValid = "db.Poem.find({ $or : [{name: 'Athanasia' }, {name: 'Portia' }]})";
Example 107. CLI syntax sort and limit results alternatives
String nativeQuery = "db.Poem.find({ '$query': { 'author': 'Oscar Wilde' }, '$orderby' : { 'name' : 1 } })";

// Using hibernate session
List<Poem> result = session.createNativeQuery( nativeQuery )
    .addEntity( Poem.class )
    .setMaxResults( 2 )
    .list();

// Using JPA entity manager
List<Poem> results = em.createNativeQuery( nativeQuery, Poem.class )
    .setMaxResults( 2 )
    .getResultList();
Example 108. CLI syntax update examples
String updateQuery = "db.Poem.findAndModify({ 'query': {'_id': 1}, 'update': { '$set': { 'author': 'Oscar Wilde' } }, 'new': true })";
List<Poem> updated = session.createNativeQuery( updateQuery ).addEntity( Poem.class ).list();

String insertQuery = "db.Poem.insert({ '_id': { '$numberLong': '11' }, 'author': 'Oscar Wilder', 'name': 'The one and wildest', 'rating': '1' } )";
int inserted = session.createNativeQuery( insertQuery ).executeUpdate();

String removeQuery = "db.Poem.remove({ '_id': { '$numberLong': '11' } })";
int removed = session.createNativeQuery( removeQuery ).executeUpdate();

Support for the $regexp operator is limited to the string syntax. We do not support the /pattern/ syntax as it is not currently supported by the MongoDB Java driver.

// Valid syntax
String nativeQuery = "{ $query : { author : { $regex : '^Oscar' } }, $orderby : { name : 1 } }";
List<Poem> result = session.createNativeQuery( nativeQuery ).addEntity( Poem.class ).list();

You can index your entities using Hibernate Search. That way, a set of secondary indexes independent of MongoDB is maintained by Hibernate Search and you can run Lucene queries on top of them. The benefit of this approach is a nice integration at the JPA / Hibernate API level (managed entities are returned by the queries). The drawback is that you need to store the Lucene indexes somewhere (file system, infinispan grid, etc). Have a look at the Infinispan section (Storing a Lucene index in Infinispan) for more info on how to use Hibernate Search.

11. Neo4j

Neo4j is a robust (fully ACID) transactional property graph database. This kind of databases are suited for those type of problems that can be represented with a graph like social relationships or road maps for example.

Hibernate OGM can connect to an embedded Neo4j or a remote server. The connection to the remote server can occur via the Bolt protocol or the HTTP API.

11.1. How to add Neo4j integration

1. Add the dependencies to your project

If your project uses Maven you can add this to the pom.xml:

<dependency>
    <groupId>org.hibernate.ogm</groupId>
    <artifactId>hibernate-ogm-neo4j</artifactId>
    <version>5.2.0.Beta1</version>
</dependency>

Alternatively you can find the required libraries in the distribution package on SourceForge

2. Use the following properties

If you want to use Neo4j in embedded mode:

hibernate.ogm.datastore.provider = neo4j_embedded
hibernate.ogm.neo4j.database_path = C:\example\mydb

If you want to connect to a remote server without authentication on localhost, you can use:

hibernate.ogm.datastore.provider = neo4j_bolt

or

hibernate.ogm.datastore.provider = neo4j_http

The difference between neo4j_bolt and neo4j_http is that in the first case Hibernate OGM will connect to the server using the Bolt protocol, in the second case it will use the HTTP interface.

You can select a different host, port and enable authentication with the following additional properties:

hibernate.ogm.datastore.provider = neo4j_bolt # or neo4j_http
hibernate.ogm.datastore.host = example.com:8989
hibernate.ogm.datastore.username = example_username
hibernate.ogm.datastore.password = example_password

The Neo4j dialects don’t require Hibernate Search to run JPQL or HQL queries.

11.2. Configuring Neo4j

The following properties are available to configure Neo4j support:

Neo4j datastore configuration properties
hibernate.ogm.neo4j.database_path

The absolute path representing the location of the Neo4j database. Example: C:\neo4jdb\mydb

hibernate.ogm.neo4j.configuration_resource_name (optional)

Location of the Neo4j embedded properties file. It can be an URL, name of a classpath resource or file system path.

hibernate.schema_update.unique_constraint_strategy (optional)

If set to SKIP, Hibernate OGM won’t create any unique constraints on the nodes representing the entities. This property won’t affect the unique constraints generated for sequences. Other possible values (defined on the org.hibernate.tool.hbm2ddl.UniqueConstraintSchemaUpdateStrategy enum) are DROP_RECREATE_QUIETLY and RECREATE_QUIETLY but the effect will be the same (since Neo4j constraints don’t have a name): keep the existing constraints and create the missing one. Default value is DROP_RECREATE_QUIETLY.

hibernate.ogm.datastore.host (optional)

The host name and the port to use when connecting to a remote server. Default value for HTTP is localhost:7474. Default value for Bolt is localhost:7687.

hibernate.ogm.datastore.username (optional)

The authentication username for the remote server. Hibernate OGM will try to authenticate only if this property is set.

hibernate.ogm.datastore.password (optional)

The password to use when connecting remotely.

When bootstrapping a session factory or entity manager factory programmatically, you should use the constants defined on org.hibernate.ogm.datastore.neo4j.Neo4jProperties when specifying the configuration properties listed above.

Common properties shared between stores are declared on OgmProperties (a super interface of Neo4jProperties).

For maximum portability between stores, use the most generic interface possible.

11.3. Storage principles

Hibernate OGM tries to make the mapping to the underlying datastore as natural as possible so that third party applications not using Hibernate OGM can still read and update the same datastore.

To make things simple, each entity is represented by a node, each embedded object is also represented by a node. Links between entities (whether to-one to to-many associations) are represented by relationships between nodes. Entity and embedded nodes are labelled ENTITY and EMBEDDED respectively.

11.3.1. Properties and built-in types

Each entity is represented by a node. Each property or more precisely column is represented by an attribute of this node.

The following types (and corresponding primitives) get passed to Neo4j without any conversion:

  • java.lang.Boolean; Optionally the annotations @Type(type = "true_false"), @Type(type = "yes_no") and @Type(type = "numeric_boolean") can be used to map boolean properties to the characters 'T'/'F', 'Y'/'N' or the int values 0/1, respectively.

  • java.lang.Character

  • java.lang.Byte

  • java.lang.Short

  • java.lang.Integer

  • java.lang.Long

  • java.lang.Float

  • java.lang.Double

  • java.lang.String

The following types get converted into java.lang.String:

  • java.math.BigDecimal

  • java.math.BigInteger

  • java.util.Calendar

    stored as `String` with the format "yyyy/MM/dd HH:mm:ss:SSS Z"
  • java.util.Date

    stored as `String` with the format "yyyy/MM/dd HH:mm:ss:SSS Z"
  • java.util.UUID

  • java.util.URL

Hibernate OGM doesn’t store null values in Neo4J, setting a value to null is the same as removing the corresponding entry from Neo4J.

This can have consequences when it comes to queries on null value.

11.3.2. Entities

Entities are stored as Neo4j nodes, which means each entity property will be translated into a property of the node. The name of the table mapping the entity is used as label.

You can use the name property of the @Table and @Column annotations to rename the label and the node’s properties.

An additional label ENTITY is added to the node.

Example 109. Default JPA mapping for an entity
@Entity
public class News {

    @Id
    private String id;
    private String title;

    // getters, setters ...
}
neo4j single node example
Example 110. Rename node label and properties using @Table and @Column
@Entity
@Table(name="ARTICLE")
public class News {

    @Id
    private String id;

    @Column(name = "headline")
    private String title;

    // getters, setters ...
}
neo4j @Column @Table example
Identifiers and unique constraints

Neo4j does not support constraints on more than one property. For this reason, Hibernate OGM will create a unique constraint ONLY when it spans a single property and it will ignore the ones spanning multiple properties.

The lack of unique constraints on node properties might result in the creation of multiple nodes with the same identifier.

Hibernate OGM will create unique constraints for the identifier of entities and for the properties annotated with:

  • @Id

  • @EmbeddedId

  • @NaturalId

  • @Column( unique = true )

  • @Table( uniqueConstraints = @UniqueConstraint(columnNames = { "column_name" } ) )

Embedded identifiers are currently stored as dot separated properties.

Example 111. Entity with @EmbeddedId
@Entity
public class News {

    @EmbeddedId
    private NewsID newsId;

    private String content

    // getters, setters ...
}

@Embeddable
public class NewsID implements Serializable {

    private String title;
    private String author;

    // getters, setters ...
}
neo4j @EmbeddedId example
Embedded objects and collections

Embedded elements are stored as separate nodes labeled with EMBEDDED.

The type of the relationship that connects the entity node to the embedded node is the attribute name representing the embedded in the java class.

Example 112. Embedded object
@Entity
public class News {

    @EmbeddedId
    private NewsID newsId;

    @Embedded
    private NewsPaper paper;

    // getters, setters ...
}

@Embeddable
public class NewsID implements Serializable {

    private String title;
    private String author;

    // getters, setters ...
}

@Embeddable
public class NewsPaper {

    private String name;
    private String owner;

    // getters, setters ...
}
neo4j @Embedded example
Example 113. @ElementCollection
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
neo4j @ElementCollection example

Note that in the previous examples no property is added to the relationships; in the following one, one property is added to keep track of the order of the elements in the list.

Example 114. @ElementCollection with @OrderColumn
@Entity
public class GrandMother {

    @Id
    private String id;

    @ElementCollection
    @OrderColumn( name = "birth_order" )
    private List<GrandChild> grandChildren = new ArrayList<GrandChild>();

    // getters, setters ...
}

@Embeddable
public class GrandChild {

    private String name;

    // getters, setters ...
}
neo4j @ElementCollection @OrderColumn example

It’s also possible to use embeddeds in a Map like in the following example:

Example 115. @ElementCollection with Map of Embedded
@Entity
public class ForumUser {

    @Id
    private String name;

    @ElementCollection
    private Map<String, JiraIssue> issues = new HashMap<>();

    // getters, setters ...
}

@Embeddable
public class JiraIssue {

    private String project;

    private Integer number;

    // getters, setters ...
}
neo4j @ElementCollection map example

This mapping has some problems, though. It will create an additional node for each element in the map and the embedded values will be stored in a node with only the EMBEDDED label. We don’t think this is a natural mapping and expect it to change in the upcoming releases.

11.3.3. Associations

An association, bidirectional or unidirectional, is always mapped using one relationship, beginning at the owning side of the association. This is possible because in Neo4j relationships can be navigated in both directions.

The type of the relationships depends on the type of the association, but in general it is the role of the association on the main side. The only property stored on the relationship is going to be the index of the association when required, for example when the association is annotated with @OrderColumn or when a java.util.Map is used.

In Neo4j nodes are connected via relationship, this means that we don’t need to create properties which store foreign column keys. This means that annotation like @JoinColumn won’t have any effect.

Example 116. Unidirectional one-to-one
@Entity
public class Vehicule {

    @Id
    private String id;
    private String brand;

    // getters, setters ...
}


@Entity
public class Wheel {

    @Id
    private String id;
    private String company;
    private double diameter;

    @OneToOne
    private Vehicule vehicule;

    // getters, setters ...
}
neo4j uni one to one example
Example 117. Bidirectional one-to-one
@Entity
public class Husband {

    @Id
    private String id;
    private String name;

    @OneToOne
    private Wife wife;

    // getters, setters ...
}

@Entity
public class Wife {

    @Id
    private String id;
    private String name;

    @OneToOne(mappedBy = "wife")
    private Husband husband;

    // getters, setters ...
}
neo4j bi one to one example
Example 118. Unidirectional one-to-many
@Entity
public class Basket {

    @Id
    private String id;

    private String owner;

    @OneToMany
    private List<Product> products = new ArrayList<Product>();

    // getters, setters ...
}

@Entity
public class Product {

    @Id
    private String name;

    private String description;

    // getters, setters ...
}
neo4j uni one to many example
Example 119. Unidirectional one-to-many using maps with defaults
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    private Map<String, Address> addresses = new HashMap<String, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}
neo4j uni one to many with map example
Example 120. Unidirectional one-to-many using maps with @MapKeyColumn
@Entity
public class User {

    @Id
    private String id;

    @OneToMany
    @MapKeyColumn(name = "addressType")
    private Map<String, Address> addresses = new HashMap<String, Address>();

    // getters, setters ...
}

@Entity
public class Address {

    @Id
    private String id;
    private String city;

    // getters, setters ...
}
neo4j uni one to many with @MapKeyColumn example
Example 121. Unidirectional many-to-one
@Entity
public class JavaUserGroup {

    @Id
    private String jug_id;
    private String name;

    // getters, setters ...
}

@Entity
public class Member {

    @Id
    private String id;
    private String name;

    @ManyToOne
    private JavaUserGroup memberOf;

    // getters, setters ...
}
neo4j uni many to one example
Example 122. Bidirectional many-to-one
@Entity
public class SalesForce {

    @Id
    private String id;
    private String corporation;

    @OneToMany(mappedBy = "salesForce")
    private Set<SalesGuy> salesGuys = new HashSet<SalesGuy>();

    // getters, setters ...
}

@Entity
public class SalesGuy {
    private String id;
    private String name;

    @ManyToOne
    private SalesForce salesForce;

    // getters, setters ...
}
neo4j bi many to one example
Example 123. Unidirectional many-to-many
@Entity
public class Student {

    @Id
    private String id;
    private String name;

    // getters, setters ...
}

@Entity
public class ClassRoom {

    @Id
    private long id;
    private String lesson;

    @ManyToMany
    private List<Student> students = new ArrayList<Student>();

    // getters, setters ...
}
neo4j uni many to many example
Example 124. Bidirectional many-to-many
@Entity
public class AccountOwner {

    @Id
    private String id;

    private String SSN;

    @ManyToMany
    private Set<BankAccount> bankAccounts;

    // getters, setters ...
}

@Entity
public class BankAccount {

    @Id
    private String id;

    private String accountNumber;

    @ManyToMany( mappedBy = "bankAccounts" )
    private Set<AccountOwner> owners = new HashSet<AccountOwner>();

    // getters, setters ...
}
neo4j bi many to many example

11.3.4. Auto-generated Values

Hibernate OGM supports the table generation strategy as well as the sequence generation strategy with Neo4j. It is generally recommended to work with the latter, as it allows a slightly more efficient querying for the next sequence value.

Sequence-based generators are represented by nodes in the following form:

Example 125. GenerationType.SEQUENCE
@Entity
public class Song {

    ...

    @Id
    @GeneratedValue( strategy = GenerationType.SEQUENCE, generator = "songSequenceGenerator" )
    @SequenceGenerator(
            name = "songSequenceGenerator",
            sequenceName = "song_sequence",
            initialValue = INITIAL_VALUE,
            allocationSize = 10)
    public Long getId() {
        return id;
    }

    ...
neo4j sequence example

Each sequence generator node is labelled with SEQUENCE. The sequence name can be specified via @SequenceGenerator#sequenceName(). A unique constraint is applied to the property sequence_name in order to ensure uniqueness of sequences.

If required, you can set the initial value of a sequence and the increment size via @SequenceGenerator#initialValue() and @SequenceGenerator#allocationSize(), respectively. The options @SequenceGenerator#catalog() and @SequenceGenerator#schema() are not supported.

Table-based generators are represented by nodes in the following form:

Example 126. GenerationType.TABLE
@Entity
public class Video {

    ...

    @Id
    @GeneratedValue( strategy = GenerationType.TABLE, generator = "video" )
    @TableGenerator(
         name = "video",
         table = "Sequences",
         pkColumnName = "key",
         pkColumnValue = "video",
         valueColumnName = "seed"
    )
    public Integer getId() {
        return id;
    }

    ...
neo4j table based sequence example

Each table generator node is labelled with TABLE_BASED_SEQUENCE and the table name as specified via @TableGenerator#table(). The sequence name is to be given via @TableGenerator#pkColumnValue(). The node properties holding the sequence name and value can be configured via @TableGenerator#pkColumnName() and @TableGenerator#valueColumnName(), respectively. A unique constraint is applied to the property sequence_name to avoid the same sequence name is used twice within the same "table".

If required, you can set the initial value of a sequence and the increment size via @TableGenerator#initialValue() and @TableGenerator#allocationSize(), respectively. The options @TableGenerator#catalog(), @TableGenerator#schema(), @TableGenerator#uniqueConstraints() and @TableGenerator#indexes() are not supported.

11.3.5. Labels summary

The maximum number of labels the database can contain is roughly 2 billion.

The following summary will help you to keep track of the labels assigned to a new node:

Table 55. Summary of the labels assigned to a new node
NODE TYPE LABELS

Entity

ENTITY, <Entity class name>

Embeddable

EMBEDDED, <Embeddable class name>

GenerationType.SEQUENCE

SEQUENCE

GenerationType.TABLE

TABLE_BASED_SEQUENCE, <Table name>

11.4. Transactions

In Neo4j, operations must be executed inside a transaction. Make sure your interactions with Hibernate OGM are within a transaction when you target Neo4J.

Example 127. Example of starting and committing transactions
Session session = factory.openSession();
Transaction tx = session.beginTransaction();

Account account = new Account();
account.setLogin( "myAccount" );
session.persist( account );

tx.commit();

...

tx = session.beginTransaction();
Account savedAccount =  (Account) session.get( Account.class, account.getId() );
tx.commit();

In the case of JTA, Hibernate OGM attaches the Neo4J internal transaction to the JTA transaction lifecycle. That way when the JTA transaction is committed or rollbacked (for example by an EJB CMT or explicitly), the Neo4J transaction is also committed or rollbacked. This makes for a nice integration in a Java EE container.

This is NOT a true JTA/XA integration but more a lifecycle alignment: changes on more than one datasource won’t be executed as a single atomic transaction.

In particular, if the JTA transaction involves multiple resources, Neo4j might commit before a failure of another resource. In this case, Neo4j won’t be able to rollback even if the JTA transaction will.

11.5. Queries

You can express queries in a few different ways:

  • using JPQL

  • using the Cypher query language

While you can use JPQL for simple queries, you might hit limitations. The current recommended approach is to use native Cypher queries if your query involves nested (list of) elements.

You don’t need Hibernate Search on the classpath to run queries.

11.5.1. JPQL queries

Hibernate OGM is a work in progress, so only a sub-set of JPQL constructs is available when using the JPQL query support. This includes:

  • simple comparisons using "<", "<=", "=", ">=" and ">"

  • IS NULL and IS NOT NULL

  • the boolean operators AND, OR, NOT

  • LIKE, IN and BETWEEN

  • ORDER BY

  • inner JOIN on embedded collections

  • projections of regular and embedded properties

Queries using these constructs will be transformed into equivalent Cypher queries.

Let us know by opening an issue or sending an email what query you wish to execute. Expanding our support in this area is high on our priority list.

11.5.2. Cypher queries

Hibernate OGM also supports Cypher queries for Neo4j. You can execute Cypher queries as shown in the following example:

Example 128. Using the JPA API
@Entity
public class Poem {

    @Id
    private Long id;

    private String name;

    private String author;

   // getters, setters ...

}

...

javax.persistence.EntityManager em = ...

// a single result query
String query1 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) RETURN n";
Poem poem = (Poem) em.createNativeQuery( query1, Poem.class ).getSingleResult();

// query with order by
String query2 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) " +
                "RETURN n ORDER BY n.name";
List<Poem> poems = em.createNativeQuery( query2, Poem.class ).getResultList();

// query with projections
String query3 = MATCH ( n:Poem ) RETURN n.name, n.author ORDER BY n.name";
List<Object[]> poemNames = (List<Object[]>) em.createNativeQuery( query3 )
                               .getResultList();

The result of a query is a managed entity (or a list thereof) or a projection of attributes in form of an object array, just like you would get from a JPQL query.

Example 129. Using the Hibernate native API
OgmSession session = ...

String query1 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) " +
                "RETURN n";
Poem poem = session.createNativeQuery( query1 )
                      .addEntity( "Poem", Poem.class )
                      .uniqueResult();

String query2 = "MATCH ( n:Poem { name:'Portia', author:'Oscar Wilde' } ) " +
                "RETURN n ORDER BY n.name";
List<Poem> poems = session.createNativeQuery( query2 )
                      .addEntity( "Poem", Poem.class )
                      .list();

Native queries can also be created using the @NamedNativeQuery annotation:

Example 130. Using @NamedNativeQuery
@Entity
@NamedNativeQuery(
   name = "AthanasiaPoem",
   query = "MATCH ( n:Poem { name:'Athanasia', author:'Oscar Wilde' } ) RETURN n",
   resultClass = Poem.class )
public class Poem { ... }

...

// Using the EntityManager
Poem poem1 = (Poem) em.createNamedQuery( "AthanasiaPoem" )
                     .getSingleResult();

// Using the Session
Poem poem2 = (Poem) session.getNamedQuery( "AthanasiaPoem" )
                     .uniqueResult();

Hibernate OGM stores data in a natural way so you can still execute queries using your favorite tool, the main drawback is that the results are going to be raw Neo4j elements and not managed entities.