JPA - Increment a numeric field through a sequence programmatically

Question

I have a JPA 2 web application (Struts 2, Hibernate 4 as JPA implementation only).

The current requirement is to add a (non-id) numeric sequential field, filled for certain rows only, to an existing entity. When inserting a new row, based on a certain condition, I need to set the new field to its highest value + 1 or to NULL.

For example:

ID     NEW_FIELD     DESCRIPTION
--------------------------------
1          1           bla bla
2                      bla bla       <--- unmatched: not needed here
3                      bla bla       <--- unmatched: not needed here
4          2           bla bla
5          3           bla bla
6          4           bla bla
7                      bla bla       <--- unmatched: not needed here
8          5           bla bla
9                      bla bla       <--- unmatched: not needed here
10         6           bla bla

In the good old SQL, it would be something like:

INSERT INTO myTable (
    id, 
    new_field, 
    description
) VALUES (
    myIdSequence.nextVal, 
    (CASE myCondition
        WHEN true 
        THEN myNewFieldSequence.nextVal
        ELSE NULL
    END),
    'Lorem Ipsum and so on....'
)

But I've no clue on how to achieve it with JPA 2.

I know I can define callbacks methods, but JSR-000317 Persistence Specification for Eval 2.0 Eval discourages some specific operations from inside it:

3.5 Entity Listeners and Callback Methods
<sub>- Lifecycle callbacks can invoke JNDI, JDBC, JMS, and enterprise beans.
- In general, the lifecycle method of a portable application should not invoke EntityManager or Query operations, access other entity instances, or modify relationships within the same persistence context.^[43] A lifecycle callback method may modify the non-relationship state of the entity on which it is invoked.</sub>

^[43] _{The semantics of such operations may be standardized in a future release of this specification.}

Summarizing, yes to JDBC (!) and EJB, no to EntityManager and other Entities.

---

EDIT

I'm trying to achieve the solution described in the answer from @anttix, but I'm encoutering some problem, so please correct me where I'm wrong.

Table

MyTable
-------------------------
ID            number (PK)
NEW_FIELD     number
DESCRIPTION   text

Main Entity

@Entity
@Table(name="MyTable")
public class MyEntity implements Serializable {

    @Id
    @SequenceGenerator(name="seq_id", sequenceName="seq_id", allocationSize=1)
    @GeneratedValue(strategy=GenerationType.SEQUENCE, generator="seq_id")
    private Long id;

    @OneToOne(cascade= CascadeType.PERSIST) 
    private FooSequence newField;

    private String description

    /* Getters and Setters */
}

Sub entity

@Entity
public class FooSequence {

    @Id
    @SequenceGenerator(name="seq_foo", sequenceName="seq_foo", allocationSize=1)
    @GeneratedValue(strategy=GenerationType.SEQUENCE, generator="seq_foo")
    private Long value;

    /* Getter and Setter */
}

DAO

myEntity.setNewField(new FooSequence());
entityManager.persist(myEntity);

Exception

Caused by: javax.transaction.RollbackException: ARJUNA016053: Could not commit transaction.

[...]

Caused by: javax.persistence.PersistenceException: org.hibernate.exception.SQLGrammarException: ERROR: relation "new_field" does not exist

[...]

Caused by: org.hibernate.exception.SQLGrammarException: ERROR: relation "new_field" does not exist

[...]

Caused by: org.postgresql.util.PSQLException: ERROR: relation "new_field" does not exist

What am I doing wrong ? I'm pretty new to JPA 2 and I've never used an entity not associated to a physical table... this approach is totally new to me.

I guess I need to put the @Column definition somewhere: how could JPA possibly know that the newField column (mapped through ImprovedNamingStrategy to new_field on the database) is retrieved through the value property of the FooSequence entity ?

Some pieces of the puzzle are missing.

---

EDIT

As asked in comments, this is the persistence.xml:

<?xml version="1.0" encoding="UTF-8"?>
<persistence version="2.0" 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">

    <persistence-unit name="MyService"  transaction-type="JTA">

        <jta-data-source>java:jboss/datasources/myDS</jta-data-source>      

        <properties>             

            <property name="hibernate.dialect" 
                     value="org.hibernate.dialect.PostgreSQLDialect" />

            <property name="hibernate.ejb.naming_strategy" 
                     value="org.hibernate.cfg.ImprovedNamingStrategy"/>

            <property name="hibernate.query.substitutions" 
                     value="true 'Y', false 'N'"/>           

         <property name="hibernate.show_sql" value="true" />
         <property name="format_sql"         value="true" />
         <property name="use_sql_comments"   value="true" />

        </properties>

    </persistence-unit>
</persistence>

Answer 1

One possible solution is to use a separate entity with its own table that will encapsulate only the new field and have an OneToOne mapping with that entity. You will then instantiate the new entity only when you encounter an object that needs the additional sequence number. You can then use any generator strategy to populate it.

@Entity
public class FooSequence {
    @Id
    @GeneratedValue(...)
    private Long value;
}

@Entity 
public class Whatever {
    @OneToOne(...)
    private FooSequnce newColumn;
}

See:

• Hibernate JPA Sequence (non-Id)
• https://forum.hibernate.org/viewtopic.php?p=2405140

A gradle 1.11 runnable SSCCE (using Spring Boot):

src/main/java/JpaMultikeyDemo.java

import java.util.List;
import javax.persistence.*;
import lombok.Data;

import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
import org.springframework.context.ConfigurableApplicationContext;
import org.springframework.context.annotation.Configuration;
import org.springframework.transaction.annotation.EnableTransactionManagement;
import org.springframework.transaction.annotation.Transactional;

@Configuration
@EnableTransactionManagement
@EnableAutoConfiguration
public class JpaMultikeyDemo {
    @Entity @Data
    public static class FooSequence {
        @Id @GeneratedValue private Long value;
    }

    @Entity @Data
    public static class FooEntity {
        @Id @GeneratedValue private Long id;
        @OneToOne           private FooSequence sequence;
    }

    @PersistenceContext
    EntityManager em;

    @Transactional
    public void runInserts() {
        // Create ten objects, half with a sequence value
        for(int i = 0; i < 10; i++) {
            FooEntity e1 = new FooEntity();
            if(i % 2 == 0) {
                FooSequence s1 = new FooSequence();
                em.persist(s1);
                e1.setSequence(s1);
            }
            em.persist(e1);
        }
    }

    public void showAll() {
        String q = "SELECT e FROM JpaMultikeyDemo$FooEntity e";
        for(FooEntity e: em.createQuery(q, FooEntity.class).getResultList())
            System.out.println(e);
    }

    public static void main(String[] args) {
        ConfigurableApplicationContext context = SpringApplication.run(JpaMultikeyDemo.class);
        context.getBean(JpaMultikeyDemo.class).runInserts();
        context.getBean(JpaMultikeyDemo.class).showAll();
        context.close();
    }
}

build.gradle

apply plugin: 'java'
defaultTasks 'execute'

repositories {
    mavenCentral()
    maven { url "http://repo.spring.io/libs-milestone" }
}

dependencies {
    compile "org.springframework.boot:spring-boot-starter-data-jpa:1.0.0.RC5"
    compile "org.projectlombok:lombok:1.12.6"
    compile "com.h2database:h2:1.3.175"
}

task execute(type:JavaExec) {
    main = "JpaMultikeyDemo"
    classpath = sourceSets.main.runtimeClasspath
}

See also: http://docs.spring.io/spring-boot/docs/current-SNAPSHOT/reference/htmlsingle/#boot-features-configure-datasource

Answer 2

This looks like it could be a good case for some AOP. First start by creating a custom field annotation @CustomSequenceGeneratedValue, and then annotate the field on the entity with it:

public class MyEntity {
...
    @CustomSequenceGeneratedValue
    private Long generatedValue;

    public void setGeneratedValue(long generatedValue) {

    }
}

Then an aspect is created to increment generated values:

@Aspect
public class CustomSequenceGeneratedValueAspect {

    @PersistenceContext 
    private EntityManager em;

    @Before("execution(* com.yourpackage.dao.SomeDao.*.*(..))")
    public void beforeSaving(JoinPoint jp) throws Throwable {
        Object[] args = jp.getArgs();
        MethodSignature ms = (MethodSignature) jp.getSignature();
        Method m = ms.getMethod();

        Annotation[][] parameterAnnotations = m.getParameterAnnotations();

        for (int i = 0; i < parameterAnnotations.length; i++) {
            Annotation[] annotations = parameterAnnotations[i];
            for (Annotation annotation : annotations) {
                if (annotation.annotationType() == CustomSequenceGeneratedEntity.class) {
                       ... find generated properties run query and call setter ...

                      ... Query query = em.createNativeQuery("select MY_SEQUENCE.NEXTVAL from dual");
                }
            }
        }
    } 
}

Then the aspect is scanned with <aop:aspectj-autoproxy />, and applied to any Spring DAO saving entities of this type. The aspect would populate the sequence generated values based on a sequence, in a transparent way for the user.

Answer 3

You mentioned being open to using JDBC. Here is how you can you use Entity Callback with JdbcTemplate, the example uses Postgres's syntax for selecting next value in a sequence, just update it to use the right syntax for your DB.

Add this to your entity class:

@javax.persistence.EntityListeners(com.example.MyEntityListener.class)

And here is listener implementation (@Qualifier and required = true are necessary for it to work):

package com.example;

import javax.persistence.PostPersist;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.beans.factory.annotation.Qualifier;
import org.springframework.jdbc.core.JdbcTemplate;
import org.springframework.stereotype.Component;
import org.springframework.transaction.annotation.Transactional;

@Component
public class MyEntityListener {

    private static JdbcTemplate jdbcTemplate;

    @Autowired(required = true)
    @Qualifier("jdbcTemplate")
    public void setJdbcTemplate(JdbcTemplate bean) {
        jdbcTemplate = bean;
    }

    @PostPersist
    @Transactional
    public void postPersis(MyEntity entity) {
        if(isUpdateNeeded(entity)) { 
            entity.setMyField(jdbcTemplate.queryForObject("select nextval('not_hibernate_sequence')", Long.class));
        }
    }

    private boolean isUpdateNeeded(MyEntity entity) {
        // TODO - implement logic to determine whether to do an update
        return false;
    }
}

Answer 4

The hacky solution I used to keep it simple is the following:

MyEntity myEntity = new MyEntity();
myEntity.setDescription("blabla");
em.persist(myEntity);
em.flush(myEntity);
myEntity.setNewField(getFooSequence());

The complete code ("pseudo-code", I've written it directly on SO so it could have typos) with transaction handling would be like :

Entity

@Entity
@Table(name="MyTable")
public class MyEntity implements Serializable {

    @Id
    @SequenceGenerator(name="seq_id", sequenceName="seq_id", allocationSize=1)
    @GeneratedValue(strategy=GenerationType.SEQUENCE, generator="seq_id")
    private Long id;

    private Long newField; // the optional sequence
    private String description
    /* Getters and Setters */
}

Main EJB:

@Stateless
@TransactionManagement(TransactionManagementType.CONTAINER) // default
public class MainEjb implements MainEjbLocalInterface {

    @Inject 
    DaoEjbLocalInterface dao;

    // Create new session, no OSIV here
    @TransactionAttribute(TransactionAttributeType.REQUIRES_NEW) 
    public Long insertMyEntity(boolean myCondition) throws Exception {

        try {
            MyEntity myEntity = dao.insertMyEntity(); 
            // if this break, no FooSequence will be generated

            doOtherStuff();
            // Do other non-database stuff that can break here. 
            // If they break, no FooSequence will be generated, 
            // and no myEntity will be persisted.                                

            if (myCondition) {
                myEntity.setNewField(dao.getFooSequence());
                // This can't break (it would have break before). 
                // But even if it breaks, no FooSequence will be generated,
                // and no myEntity will be persisted.
            }
        } catch (Exception e){
            getContext().setRollbackOnly();
            log.error(e.getMessage(),e);
            throw new MyException(e);
        }    
    }
}

DAO EJB

@Stateless
@TransactionManagement(TransactionManagementType.CONTAINER) // default
public class DaoEjb implements DaoEjbLocalInterface {

    @PersistenceContext( unitName="myPersistenceUnit")
    EntityManager em;

    // default, use caller (MainEJB) session
    @TransactionAttribute(TransactionAttributeType.REQUIRED) 
    public MyEntity insertMyEntity() throws Exception{
        MyEntity myEntity = new MyEntity();
        myEntity.setDescription("blabla");
        em.persist(myEntity);
        em.flush(); // here it will break in case of database errors, 
                    // eg. description value too long for the column.
                    // Not yet committed, but already "tested".
        return myEntity;
    }

    // default, use caller (MainEJB) session
    @TransactionAttribute(TransactionAttributeType.REQUIRED) 
    public Long getFooSequence() throws Exception {
        Query query = em.createNativeQuery("SELECT nextval('seq_foo')");
        return ((BigInteger) query.getResultList().get(0)).longValue();
    }
}

This will guarantee there will be no gaps in the FooSequence generation.

The only drawback, that I don't care at all in my use case, is that FooSequence and the @Id sequence are not synchronized, so two concurrent inserts may have "inverted" FooSequence values, respecto to their order of arrive, eg.

ID  NEW FIELD
-------------
 1      2
 2      1