6

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
- 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.

[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> 
20
  • @AndreiI Please, can you elaborate ? I don't get it. Which is the old column ? Table was ID - DESCRIPTION, entity was Id - description, now I added a non-id, non-pk field to both the table and the entity, and need to assign it a value only sometimes. Commented Mar 21, 2014 at 16:24
  • Add @JoinColumn(name="INCR_FIELD") to your private FooSequence newField;, and add the FooSequence class to your persistence.xml (if needed). Afterwards, in order to get the value of NEW_FIELD use getFooSequence().getId() Commented Mar 21, 2014 at 16:33
  • No heck, same error (I added @JoinColumn(name="new_field") to private FooSequence newField; , that was the only change I made). Please, post it as an answer with some more details, I'll assign the bounty as soon as it will work, no matter the ending period, and I'll upvote every helping answer immediately. Commented Mar 21, 2014 at 16:56
  • please try using a different column than new_field (to be sure it is not caused by that column), that's why I used a new INCR_FIELD column. Commented Mar 21, 2014 at 16:58
  • 1
    @AndreaLigios exactly: it works with two tables, each with an ID field, but I doubt that is a problem for you. In order to take get the NEW_FIELD in your code, you should call getFooSequence().getId() (as I mentioned in one of my comments above). Commented Mar 27, 2014 at 11:26

4 Answers 4

4
+50

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:

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

Sign up to request clarification or add additional context in comments.

13 Comments

Thanks, I'd found that answer before asking my question, but wasn't sure if it was the best way to go. I've started trying today, but I'm encountering some errors, that I'll add to the question; meanwhile, If you have implemented (or seen) the solution you described, could you please expand its usage description ? Since I've always used Entities bound to DB tables, while this one should be a column of the table bound to the Parent entity (Whatever), I'm not sure if I'm doing it properly
As said in a comment above, the problem is that FooSequence is created, but FooSequence.value is null
New objects do not get their ID fields populated until they are persisted and flushed e.g. em.persist(newObj); em.flush(); Hibernate will generate ID-s and order SQL statements before they are executed (flushed). Thus the fact that the value shows up as null initially will not stop the relation from being properly inserted/updated.
You are totally right, the null problem was a non problem, after writing that I found out the real issue was @OneToOne or @ManyToOne on references an unknown entity: FooSequence, then I've done a zillion tries and gave up using a workaround (I had few hours to experiment with this, unluckily). Can't post SQL (that was absolutely normal) right now, code is changed too much, even in the unit tests, and I've no time to clean it up to post it here; I still would like to see the above code running in an SSCCE, though :/ Thanks a lot for your time
@andrea-ligios you do realize that a JPA SSCCE is not very small and not very self-contained ... That said, fortunately we have Spring boot ;) See the updated answer.
|
3

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.

4 Comments

Thanks, I know AOP, I've used it sometimes in the past, BTW it would be a little overkill in this case; I already have alternative ways to do my job, I'm just curious about the other-entity answer, since it seems to be the right JPA way to do it, but I can't make it work :/ Thanks the same, +1.
the problem is that JPA says that @GeneratedValue cannot be used for non id fields, see in this PDF page 375, it says [97] Portable applications should not use the GeneratedValue annotation on other persistent fields or properties. code.google.com/p/jgk-spring-recipes/downloads/…
so without an official solution, i think the only way is to either do some AOP, or create a hibernate custom type. one optimization is to use a hi-Lo sequence in the database (reserves for example 100 keys in one go, so that the sequence is not called 100 times, only one)
The solution I was trying to set-up received 25 upvotes in the original question, and has been claimed to run properly, so that should be the official JPA "hack" (if not JPA "solution"), and someone must have been able tu make it work; I was trying that until one hour ago, when I dropped it and switched to an easier, 30min workaround, that I will post if noone will provide a solution. As said, your is a great solution to a more complex problem, and I'll use it in future, best fitting scenarios. I would like to avoid raw Hibernate solutions (while using Hibernate!), AspectJ is even beyond that
1

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; } } 

Comments

0

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 

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.