Question

Well the question pretty much says everything. Using JPARepository how do I update an entity?

JPARepository has only a save method, which does not tell me if it’s create or update actually. For example, I insert a simple Object to the database User, which has three fields: firstname, lastname and age:

 @Entity
 public class User {

  private String firstname;
  private String lastname;
  //Setters and getters for age omitted, but they are the same as with firstname and lastname.
  private int age;

  @Column
  public String getFirstname() {
    return firstname;
  }
  public void setFirstname(String firstname) {
    this.firstname = firstname;
  }

  @Column
  public String getLastname() {
    return lastname;
  }
  public void setLastname(String lastname) {
    this.lastname = lastname;
  }

  private long userId;

  @Id
  @GeneratedValue(strategy=GenerationType.AUTO)
  public long getUserId(){
    return this.userId;
  }

  public void setUserId(long userId){
    this.userId = userId;
  }
}

Then I simply call save(), which at this point is actually an insert into database:

 User user1 = new User();
 user1.setFirstname("john"); user1.setLastname("dew");
 user1.setAge(16);

 userService.saveUser(user1);// This call is actually using the JPARepository: userRepository.save(user);

So far so good. Now I want to update this user, say change his age. For this purpose I could use a Query, either QueryDSL or NamedQuery, whatever. But, considering I just want to use spring-data-jpa and the JPARepository, how do I tell it that instead of an insert I want to do an update?

Specifically, how do I tell spring-data-jpa that users with the same username and firstname are actually EQUAL and that the existing entity supposed to be updated? Overriding equals did not solve this problem.

Answer 1

Identity of entities is defined by their primary keys. Since firstname and lastname are not parts of the primary key, you cannot tell JPA to treat Users with the same firstnames and lastnames as equal if they have different userIds.

So, if you want to update a User identified by its firstname and lastname, you need to find that User by a query, and then change appropriate fields of the object your found. These changes will be flushed to the database automatically at the end of transaction, so that you don’t need to do anything to save these changes explicitly.

EDIT:

Perhaps I should elaborate on overall semantics of JPA. There are two main approaches to design of persistence APIs:

• insert/update approach. When you need to modify the database you should call methods of persistence API explicitly: you call insert to insert an object, or update to save new state of the object to the database.

• Unit of Work approach. In this case you have a set of objects managed by persistence library. All changes you make to these objects will be flushed to the database automatically at the end of Unit of Work (i.e. at the end of the current transaction in typical case). When you need to insert new record to the database, you make the corresponding object managed. Managed objects are identified by their primary keys, so that if you make an object with predefined primary key managed, it will be associated with the database record of the same id, and state of this object will be propagated to that record automatically.

JPA follows the latter approach. save() in Spring Data JPA is backed by merge() in plain JPA, therefore it makes your entity managed as described above. It means that calling save() on an object with predefined id will update the corresponding database record rather than insert a new one, and also explains why save() is not called create().