JPA 与共享主键的单向@OneToOne 关系始终触发辅助查询，即使 fetchType 为 EAGER

Question

I am building a blog system, and like to provide the upvote/downvote feature for the blog. Since the vote count number of blog should be persisted, i choose to use MySQL to act as the data store. And i use Spring JPA(Hibernate) to do the ORM job. Here's my data objects:

class Blog{
    // ...
    @Id
    @GeneratedValue(strategy = GenerationType.IDENTITY)
    private Long id;
    
    @OneToOne(optional = false, fetch = FetchType.EAGER)
    @PrimaryKeyJoinColumn
    private BlogVoteCounter voteCounter;    
}

And the counter class:

@Entity
public class BlogVoteCounter extends ManuallyAssignIdEntitySuperClass<Long> {
    @Id
    private Long id;

    private Integer value;
}

The reason why i separate the BlogVoteCounter from Blog is that i think the voteCount field will be modified by a totally different frequency comparing to other fields of Blog, since i want to use cache to cache the Blog , following this guide , i choose to separate them.

However, since the VoteCount field might be always needed when return the Blog object to the front end, and to avoid the n+1 problem, i declared the BlogVoteCounter field in Blog class with EAGER fetch type.

I've already seen this article . Thus according to my personal comprehension, i use unidirectional relationship and only declare OneToOne in the Blog side.

However, when i examine the query, it turns out that jpa will still trigger a secondary query to retrieve BlogVoteCounter from database without simply using a join when use findAll method on BlogRepository .

    select
        blogvoteco0_.id as id1_2_0_,
        blogvoteco0_.value as value2_2_0_ 
    from
        blog_vote_counter blogvoteco0_ 
    where
        blogvoteco0_.id=?

So how should i config, to always make the BlogVoteCounter field in Blog be fetched eagerly.

---

The usage of ManuallyAssignIdEntitySuperClass is following the Spring JPA doc , since i manually assign id for BlogVoteCounter class.

@MappedSuperclass
public abstract class ManuallyAssignIdEntitySuperClass<ID> implements Persistable<ID> {

    @Transient
    private boolean isNew = true;

    @Override
    public boolean isNew() {
        return isNew;
    }

    @PrePersist
    @PostLoad
    void markNotNew(){
        this.isNew = false;
    }
}

And the BlogRepository is derived from JpaRepository

public interface BlogRepository extends JpaRepository<Blog, Long>{
    // ...
}

I trigger the query by using findAll method, but using findById or other conditional query seems no difference.

Answer 1

When to fetch vs How to fetch : fetchType defines when to fetch the association ( instantly vs later when someone access ) the association but not how to fetch the association(ie second select vs join query). So from JPA Spec point of view, EAGER means dont wait until someone access that field to populate it but JPA provider is free to use JOIN or second select as long as they do it immediately.

Even though they are free to use join vs second select, still I thought they should have optimised for join in the case of EAGER. So interested in finding out the logical reasoning for not using the join

1. Query generated for repository.findById(blogId);

    select
        blog0_.id as id1_0_0_,
        blog0_.vote_counter_id as vote_cou2_0_0_,
        blogvoteco1_.id as id1_1_1_,
        blogvoteco1_.value as value2_1_1_ 
    from
        blog blog0_ 
    inner join
        blog_vote_counter blogvoteco1_ 
            on blog0_.vote_counter_id=blogvoteco1_.id 
    where
        blog0_.id=?

2. Updated Mapping

public class Blog {

    @Id
    private Long id;

    @ManyToOne(optional = false, cascade = ALL, fetch = FetchType.EAGER)
    @PrimaryKeyJoinColumn
    private BlogVoteCounter voteCounter;

    public Blog() {
    }

    public Blog(Long id, BlogVoteCounter voteCounter) {
        this.id = id;
        this.voteCounter = voteCounter;
    }

    public Long getId() {
        return id;
    }

    public void setId(Long id) {
        this.id = id;
    }

    public BlogVoteCounter getVoteCounter() {
        return voteCounter;
    }

    public void setVoteCounter(BlogVoteCounter voteCounter) {
        this.voteCounter = voteCounter;
    }
}

3. Issues with current Mapping

• As per your mapping, it is impossible to create blog and votecounter as it causes a chicken and egg problem. ie
• blog and votecounter need to share the same primary key
• blog's primary key is generated by database.
• so in order to get the primary key of blog and assign it to votecounter as well, you need to store blog first
• but the @OneToOne relationship is not optional, so you cannot store blog first alone

4.Changes

• Either need to make the relationship optional so blog can be stored first, get the id, assign to BlogVoteCounter and save the counter
• Or Don't auto generate Id and manually assign the id so blog and votecounter can be saved at the same time.(I have gone for this option but you can do first option)

5.Notes

• default repository.findAll was generating 2 queries so I overridden that method to generate one join query

public interface BlogRepository extends JpaRepository<Blog, Long> {

    @Override
    @Query("SELECT b from Blog b join fetch b.voteCounter ")
    List<Blog> findAll();
}

    select
        blog0_.id as id1_0_0_,
        blogvoteco1_.id as id1_1_1_,
        blog0_.vote_counter_id as vote_cou2_0_0_,
        blogvoteco1_.value as value2_1_1_ 
    from
        blog blog0_ 
    inner join
        blog_vote_counter blogvoteco1_ 
            on blog0_.vote_counter_id=blogvoteco1_.id