Spring/Hibernate optimization: JPA with @OneToOne mapping

Lately I wanted to see how Hibernate actually construct sql statements. And it seems like it makes somtimes unnecessary joins. Those are my playground's results.

Repo: https://github.com/toficzak/social-app

Setup:

	@Entity
	@Table(name = "users")
	public class User extends BaseEntity {
	}
	@Entity
	public class Post extends BaseEntity {
	@NotBlank
	String message;
	@NotNull
	@ManyToOne(fetch = FetchType.LAZY)
	@JoinColumn(name = "author_id", referencedColumnName = "id")
	User author;
	}

view raw Classes.java hosted with ❤ by GitHub

Post repository consists of 3 methods:

	@Repository
	public interface PostRepository extends CrudRepository<Post, Long> {
	Optional<Post> findTopByAuthorOrderByIdDesc(User author);
	Optional<Post> findTopByAuthorIdOrderByIdDesc(Long authorId);
	@Query("from Post where author_id = ?1 order by id desc")
	Optional<Post> findWithCustomQueryOrderByIdDesc(Long authorId);
	}

view raw PostRepository.java hosted with ❤ by GitHub

Sql generated by Hibernate:

	--> By author id:
	Hibernate:
	select
	post0_.id as id1_0_,
	post0_.author_id as author_i3_0_,
	post0_.message as message2_0_
	from
	post post0_
	left outer join
	users user1_
	on post0_.author_id=user1_.id
	where
	user1_.id=?
	order by
	post0_.id desc limit ?
	--> By author:
	Hibernate:
	select
	post0_.id as id1_0_,
	post0_.author_id as author_i3_0_,
	post0_.message as message2_0_
	from
	post post0_
	where
	post0_.author_id=?
	order by
	post0_.id desc limit ?
	--> With custom query:
	Hibernate:
	select
	post0_.id as id1_0_,
	post0_.author_id as author_i3_0_,
	post0_.message as message2_0_
	from
	post post0_
	where
	author_id=?
	order by
	post0_.id desc

view raw Hibernate.sql hosted with ❤ by GitHub

How did I measure execution time:

	startTime = System.nanoTime();
	method();
	endTime = System.nanoTime();
	duration = (endTime - startTime);
	results.add(duration);
	average = results.stream().mapToLong(Long::longValue).average().getAsDouble()

view raw MeasureMethod.java hosted with ❤ by GitHub

  

Simple select measure

By author id: 1952090.938

By author (User provided): 1048579.671

By author (User not provided): 1021450.464

By custom query: 935087.347

My custom JPA query is the fastest option - it skips whole User join section. But why is it needed at all? Seems like an error in my design.