Regression for method-based query where predicate compares primary key of `@ManyToOne` relationship

[scordio]

Given the following entities:

@Entity
class Author {

  @Id @GeneratedValue(strategy = GenerationType.AUTO)
  private Long id;

  private String name;

  // getters and setters omitted
}

@Entity
class Book {

  @Id @GeneratedValue(strategy = GenerationType.AUTO)
  private Long id;

  private String name;

  @ManyToOne(fetch = FetchType.LAZY)
  private Author author;

  // getters and setters omitted
}

and the following repository:

interface BookRepository extends Repository<Book, Long> {

  Book save(Book book);

  List<Book> findAllByAuthorId(Long authorId);

}

findAllByAuthorId seems to trigger an unnecessary JOIN.

I initially noticed this behavior on Spring Boot 2.7.14 with DB2 for z/OS so I tried to reproduce it locally on newer Spring Boot versions.

Surprisingly, 3.0.x and 3.1.x seem to work properly, i.e., no unnecessary JOIN, while the JOIN is back again on 3.2.x and 3.3.x. That's why I mentioned it as a regression in the title.

Here are the derived queries for each Spring Boot version with an H2 database, captured with P6Spy:

2.7.18

select book0_.id as id1_1_, book0_.author_id as author_i3_1_, book0_.name as name2_1_ from book book0_ left outer join author author1_ on book0_.author_id=author1_.id where author1_.id=1;

3.0.13

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.1.8

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.2.2

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 left join author a1_0 on a1_0.id=b1_0.author_id where a1_0.id=1;

3.3.0-SNAPSHOT

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 left join author a1_0 on a1_0.id=b1_0.author_id where a1_0.id=1;

Reproducers:

• 2.7.18: BookRepositoryTest
• 3.x: BookRepositoryTest (different versions covered with workflow matrix)

[christophstrobl]

thanks for reporting and the reproducer - have you tried to do the same query with plain JPA/hibernate as well?

[scordio]

I suppose you're referring to JPQL, right @christophstrobl?

I've just added the following to the reproducers:

@Query("SELECT b FROM Book b WHERE b.author.id = :authorId")
List<Book> findAllByAuthorIdWithJPQL(@Param("authorId") Long authorId);

It looks good on all versions:

2.7.18

select book0_.id as id1_1_, book0_.author_id as author_i3_1_, book0_.name as name2_1_ from book book0_ where book0_.author_id=1;

3.0.13

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.1.8

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.2.2

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.3.0-SNAPSHOT

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

[scordio]

For completeness, I also tried with JpaSpecificationExecutor and a Specification:

private static Specification<Book> byAuthorId(Long authorId) {
  return (root, query, builder) -> builder.equal(root.get("author").get("id"), authorId);
}

It also looks good on all versions:

2.7.18

select book0_.id as id1_1_, book0_.author_id as author_i3_1_, book0_.name as name2_1_ from book book0_ where book0_.author_id=1;

3.0.13

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.1.8

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.2.2

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

3.3.0-SNAPSHOT

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=1;

[quaff]

have you tried to do the same query with plain JPA/hibernate as well?

	CriteriaBuilder cb = entityManager.getCriteriaBuilder();
	CriteriaQuery<Book> cq = cb.createQuery(Book.class);
	Root<Book> root = cq.from(Book.class);
	cq.select(root).where(cb.equal(root.get("author").get("id"), homer.getId()));
	entityManager.createQuery(cq).getResultList();

will produce expected sql

select b1_0.id,b1_0.author_id,b1_0.name from book b1_0 where b1_0.author_id=?

[quaff]

@scordio It works as expected if you downgrade hibernate to 6.2.20.Final which is used by 3.1.8, I confirm it is a regression introduced by hibernate 6.4.1.Final, I reported it to hibernate team, see https://hibernate.atlassian.net/browse/HHH-17706.

[scordio]

Thanks @quaff for helping pinpoint the root cause!

Closing in favor of HHH-17706 and hibernate/hibernate-orm#7782.

[scordio]

I am reopening this issue as hibernate/hibernate-orm#7782 has been rejected.

@mbladel mentioned in HHH-17706:

I would suggest using implicit join paths (i.e. root.simpleEntity.id or root.get( "simpleEntity" ).get( "id" ) with Criteria) to take advantage of foreign-key optimization if really needed.

Is it something that could be done in Spring Data JPA?

[quaff]

Here is workaround:

interface BookRepository extends JpaRepository<Book, Long> {

    default List<Book> findAllByAuthorId(Long authorId) {
        Author author = new Author();
        author.setId(authorId);
        return findAllByAuthor(author);
    }

    List<Book> findAllByAuthor(Author author);

}

[christophstrobl]

Thank you both @scordio & @quaff - we'll see if there's anything we can do on our side.