Support for Composite Keys using `@Embeded` and `@Id`

[spring-projects-issues]

alpet opened DATAJDBC-352 and commented

Enable usage of @Embdedded and @Id together. I see two scenarios here. One would be for read only repositories and the other for the CRUD repository. With respect of the legacy applications very often read only is enough. How would Embedded participate in a query ? 

 

One reason to use a lower level persistence framework in my opinion are legacy applications who very often have Composite Keys. Emabling such feature in spring data jdbc would increase the possible users of the framework

---

13 votes, 14 watchers

[spring-projects-issues]

Lovro Pandžić commented

Hello, Jens Schauder, any idea when this might get fixed? Composite keys seem to be a pretty fundamental feature for any persistence layer library. Since there are no workarounds (I don't consider adding a new unique key to a production table because of library limitations a viable workaround), users that do use composite keys can't use Spring Data JDBC

[spring-projects-issues]

Jens Schauder commented

This is a feature we want to have. Unfortunately we currently can't see far enough in the future to tell when it will happen.

[imbananko]

hi
any updates on this?

[jiahaolinTHG]

hi
any updates on this?

[ttnsgn]

meanwhile this feature unsupported yet is there a workaround for this?

[kao78]

Hi
any update on this feature?

[porfirioribeiro]

I managed to make it work, with a workaround.
Instead of having multiple @Id you specify none
Create a class for being your primary key (I guess you can use a some generic Pair or Tuple class)
Create methods on repository to get/delete byId, because the default ones will not work

Working example in Kotlin (it would work also in a Java equivalent):

// Database table model
@Table("policy")
class Policy(
  val userId: UUID,
  val entityId: UUID,
  val name: String,
)

// Composite primary key
data class PolicyPK(
  val userId: UUID? = null,
  val entityId: UUID? = null,
) : Serializable

// Repository
interface PolicyRepository : R2dbcRepository<Policy, PolicyPK> {
  /**
   * replaces: findById(id: PolicyPk)
   * Find by id does not work because of the composite id so we query with all id fields
   */
  fun findByUserIdAndEntityId(userId: UUID, entityId: UUID): Mono<Policy>
  
  /**
   * replaces: delete(policy)
   * Delete with entity does not work because of the composite id so we delete with all id fields
   */
  fun deleteByUserIdAndEntityId(userId: UUID, entityId: UUID): Mono<Void>
}

If you really need to have the default findById and others, I guess you can override them on your repository and specify a @Query

I hope this helps someone, spend a bunch of time to figure this out

[st0ke]

Just in case if somebody wants to know @porfirioribeiro's workaround (no id property) does not work with R2dbcEntityTemplate

[xuandungdoan]

@porfirioribeiro your way also can't save any record, the application throws me "IllegalStateException: Required identifier property not found "

[porfirioribeiro]

@xuandungdoan you're right .save() also does not work
You have to make your own save function using a query
In my case i needed a upsert function so i ended up with something like this

  @Modifying
  @Query(
    """
insert into policy.policy(user_id, entity_id, resource_id, resource_type, roles, created_at, updated_at)
values (:#{#policy.userId}, :#{#policy.entityId}, :#{#policy.resourceId}, :#{#policy.resourceType}, :#{#policy.roles}, :#{#policy.createdAt}, now())
on conflict (user_id, entity_id, resource_id) do update set roles=:#{#policy.roles}, updated_at=now()"""
  )
  fun upsert(policy: Policy): Mono<Int>

It kinda sucked to have to do all this workaround to make it work, but at least I could still be using R2DBC and part of the Repository stuff, and not have to change all this back to use JPA/Hibernate

[Kshitij09]

Any updates on this? I need this feature to be able refer an aggregate across multiple roots using a pojo class representing the composite key