Spring Data MongoDB cascade save on DBRef objects

Maciej WalkowiakNovember 29th, 2013Last Updated: November 29th, 2013

2 1,083 3 minutes read

Spring Data MongoDB by default does not support cascade operations on referenced objects with @DBRef annotations as reference says:

The mapping framework does not handle cascading saves. If you change an Account object that is referenced by a Person object, you must save the Account object separately. Calling save on the Person object will not automatically save the Account objects in the property accounts.

That’s quite problematic because in order to achieve saving child objects you need to override save method in repository in parent or create additional â€œserviceâ€ methods like it is presented in here.

In this article I will show you how it can be achieved for all documents using generic implementation of AbstractMongoEventListener.

@CascadeSave annotation

Because we can’t change @DBRef annotation by adding cascade property lets create new annotation @CascadeSave that will be used to mark which fields should be saved when parent object is saved.

package pl.maciejwalkowiak.springdata.mongodb;

import java.lang.annotation.ElementType;
import java.lang.annotation.Retention;
import java.lang.annotation.RetentionPolicy;
import java.lang.annotation.Target;

@Retention(RetentionPolicy.RUNTIME)
@Target({ ElementType.FIELD })
public @interface CascadeSave {

}

CascadingMongoEventListener

Next part is to implement handler for this annotation. We will use for that powerful Spring Application Event mechanism. In particular we will extend AbstractMongoEventListener to catch saved object before it is converted to Mongoâ€™s DBObject.

How does it work? When object MongoTemplate#save method is called, before object is actually saved it is being converted into DBObject from MongoDB api. CascadingMongoEventListener implemented below provides hook that catches object before its converted and:

goes through all its fields to check if there are fields annotated with @DBRef and @CascadeSave at once.
when field is found it checks if @Id annotation is present
child object is being saved

package pl.maciejwalkowiak.springdata.mongodb;

import org.bson.types.ObjectId;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.data.annotation.Id;
import org.springframework.data.mapping.model.MappingException;
import org.springframework.data.mongodb.core.MongoOperations;
import org.springframework.data.mongodb.core.mapping.DBRef;
import org.springframework.data.mongodb.core.mapping.event.AbstractMongoEventListener;
import org.springframework.stereotype.Component;
import org.springframework.util.ReflectionUtils;

import java.lang.reflect.Field;

public class CascadingMongoEventListener extends AbstractMongoEventListener {
  @Autowired
  private MongoOperations mongoOperations;

  @Override
  public void onBeforeConvert(final Object source) {
      ReflectionUtils.doWithFields(source.getClass(), new ReflectionUtils.FieldCallback() {

          public void doWith(Field field) throws IllegalArgumentException, IllegalAccessException {
              ReflectionUtils.makeAccessible(field);

              if (field.isAnnotationPresent(DBRef.class) && field.isAnnotationPresent(CascadeSave.class)) {
                  final Object fieldValue = field.get(source);

                  DbRefFieldCallback callback = new DbRefFieldCallback();

                  ReflectionUtils.doWithFields(fieldValue.getClass(), callback);

                  if (!callback.isIdFound()) {
                      throw new MappingException("Cannot perform cascade save on child object without id set");
                  }

                  mongoOperations.save(fieldValue);
              }
          }
      });
  }

  private static class DbRefFieldCallback implements ReflectionUtils.FieldCallback {
      private boolean idFound;

      public void doWith(Field field) throws IllegalArgumentException, IllegalAccessException {
          ReflectionUtils.makeAccessible(field);

          if (field.isAnnotationPresent(Id.class)) {
              idFound = true;
          }
      }

      public boolean isIdFound() {
          return idFound;
      }
  }
}

Mapping requirements

As you can see in order to make thing work you need to follow some rules:

parent’s class child property has to be mapped with @DBRef and @CascadeSave
child class needs to have property annotated with @Id and if that id is supposed to be autogenerated it should by type of ObjectId

Usage

In order to use cascade saving in your project you need just to register CascadingMongoEventListener in Spring Context:

<bean class="pl.maciejwalkowiak.springdata.mongodb.CascadingMongoEventListener" />

Let’s test it

In order to show an example I made two document classes:

@Document
public class User {
  @Id
  private ObjectId id;
  private String name;

  @DBRef
  @CascadeSave
  private Address address;

  public User(String name) {
      this.name = name;
  }

  // ... getters, setters, equals hashcode
}

@Document
public class Address {
  @Id
  private ObjectId id;
  private String city;

  public Address(String city) {
      this.city = city;
  }

  // ... getters, setters, equals hashcode
}

In test there is one user with address created and then user is saved. Test will cover only positive scenario and its just meant to show that it actually works (applcationContext-tests.xml contains only default Spring Data MongoDB beans and CascadingMongoEventListener registered):

@RunWith(SpringJUnit4ClassRunner.class)
@ContextConfiguration(locations = {"classpath:applcationContext-tests.xml"})
public class CascadingMongoEventListenerTest {

  @Autowired
  private MongoOperations mongoOperations;

  /**
  * Clean collections before tests are executed
  */
  @Before
  public void cleanCollections() {
      mongoOperations.dropCollection(User.class);
      mongoOperations.dropCollection(Address.class);
  }

  @Test
  public void testCascadeSave() {
      // given
      User user = new User("John Smith");
      user.setAddress(new Address("London"));

      // when
      mongoOperations.save(user);

      // then
      List<User> users = mongoOperations.findAll(User.class);
      assertThat(users).hasSize(1).containsOnly(user);

      User savedUser = users.get(0);
      assertThat(savedUser.getAddress()).isNotNull().isEqualTo(user.getAddress());

      List<Address> addresses = mongoOperations.findAll(Address.class);
      assertThat(addresses).hasSize(1).containsOnly(user.getAddress());
  }
}

We can check that also in Mongo console:

> db.user.find()
{ "_id" : ObjectId("4f9d1bab1a8854250a5bf13e"), "_class" : "pl.maciejwalkowiak.springdata.mongodb.domain.User", "name" : "John Smith", "address" : { "$ref" : "address", "$id" : ObjectId("4f9d1ba41a8854250a5bf13d") } }
> db.address.find()
{ "_id" : ObjectId("4f9d1ba41a8854250a5bf13d"), "_class" : "pl.maciejwalkowiak.springdata.mongodb.domain.Address", "city" : "London" }

Summary

With this simple solution we can finally save child objects with one method call without implementing anything special for each document class.

I believe that we will find this functionality together with cascade delete as part Spring Data MongoDB release in the future. Solution presented here works but:

it requires to use additional annotation
uses reflection API to iterate through fields which is not the fastest way to do it (but feel free to implement caching if needed)

If that could be part of Spring Data MongoDB instead of additional annotation @DBRef could have additional property cascade. Instead of reflection we could use MongoMappingContext together with MongoPersistentEntity. I’ve started already to prepare pull request with those changes. We will see if it will be accepted by Spring Source team.

Reference: Spring Data MongoDB cascade save on DBRef objects from our JCG partner Maciej Walkowiak at the Software Development Journey blog.