Unwanted Cross Join in JPA CriteriaQuery select in subquery

I’m getting what I think are needless CROSS JOINs when I’m doing a select IN SUBQUERY, which is hurting performance. I’m using Postgres if that makes a difference.

I’m aiming to generate the following query

select a1.first_name from author a1
where a1.last_name = ?
  and (a1.id in
       (select distinct b.author_id
        from book b    
          where (b.published_on between ? and ?)
        group by b.author_id
        having count(b.author_id) >= 2))

But I get

select a1.first_name from author a1
where a1.last_name = ?
  and (a1.id in
       (select distinct b.author_id
        from book b
          cross join author a2 where b.author_id = a2.id -- <<< I don't want this cross join!
          and (b.published_on between ? and ?)
        group by b.author_id
        having count(b.author_id) >= 2))

Code

        CriteriaBuilder cb = entityManager.getCriteriaBuilder();
        CriteriaQuery<String> cq = cb.createQuery(Author.class);
        Root<Author> authorRoot = cq.from(Author.class);

        Subquery<Long> countSubquery = cq.subquery(Long.class);
        Root<Book> bookRoot = countSubquery.from(Book.class);
        Expression<Long> count = cb.count(bookRoot.get(Book_.author));

        countSubquery.select(bookRoot.get(Book_.AUTHOR))
            .distinct(true)
            .where(cb.between(bookRoot.get(Book_.publishedOn),
                LocalDate.of(2021, MARCH, 1),
                LocalDate.of(2021, MARCH, 31)))
            .groupBy(bookRoot.get(Book_.author))
            .having(cb.greaterThanOrEqualTo(count, 2L));

        cq.where(
            cb.equal(authorRoot.get(Author_.lastName), "Smith"),
            cb.in(authorRoot.get(Author_.ID)).value(countSubquery));

        cq.select(authorRoot.get(Author_.FIRST_NAME));

        TypedQuery<String> query = entityManager.createQuery(cq);

        return query.getResultList();

In reality I’m generating the queries from a user driven query builder, this code recreates the exact problem I’m having.

When using the query builder the user could end up with multiple select in subqueries so I need this to perform as well as possible.

I don’t see why I should need any join/cross join for my query to work.

Entities

@Entity
public class Author {

    @Id
    @GeneratedValue
    private Long id;

    private String firstName;
    private String lastName;

    @OneToMany(mappedBy = "author", cascade = CascadeType.ALL, fetch = FetchType.LAZY)
    private Set<Book> books;

}

@Entity
public class Book {

    @Id
    @GeneratedValue
    private Long id;

    private String name;

    @ManyToOne(fetch = FetchType.LAZY)
    @JoinColumn(name = "author_id")
    private Author author;

    private LocalDate publishedOn;
    
}

Answer

This expression: bookRoot.get(Book_.author) means you’re joining Author to Book implicitly.

To get rid of the extra join, you would have to either use a native query, or map Book.author_id once more as a simple column:

@Column(name = "author_id", insertable = false, updatable = false)
private Long authorId;

And use Book_.authorId instead.