Spring-Data-JPA with QueryDslPredicateExecutor and Joining into a collection

digitaljoel picture digitaljoel · Feb 7, 2014 · Viewed 24.5k times · Source

Let's say I have a data model like this (pseudocode):

@Entity
Person {
    @OneToMany
    List<PersonAttribute> attributes;
}

@Entity
PersonAttribute {
    @ManyToOne
    AttributeName attributeName;

    String attributeValue;
}

@Entity
AttributeName {
    String name;
}

I have a Spring-Data-JPA repository defined such as:

public interface PersonRepository extends PagingAndSortingRepository<Person, Long>, QueryDslPredicateExecutor<Person>{}

I see in the QueryDSL documentation that there is a mechanism to Join from the Person to the PersonAttribute, but it looks like you need access to the QueryDsl Query object, which the client of the repository wouldn't have.

What I would like to do with my Predicate is to find all those Persons that have an AttributeValue (there's one join) with a value of "blue" and an AttributeName (there's another join) with a name of "eyecolor". I'm not sure how I would do that with an any() and enforce that I only get those with eye_color=blue and not those with shoe_color=blue.

I was hoping I could do something like this:

QPerson person = QPerson.person;
QPersonAttribute attribute = person.attributes.any();

Predicate predicate = person.name.toLowerCase().startsWith("jo")
    .and(attribute.attributeName().name.toLowerCase().eq("eye color")
          .and(attribute.attributeValue.toLowerCase().eq("blue")));

but with the any() in there it just matches anything with an attribute value of "blue" and anything with an "eye color" attribute regardless of color. How I can make those conditions apply to the same attribute within the set?

Answer

Timo Westk&#228;mper picture Timo Westkämper · Feb 7, 2014

You can't directly join a column in a predicate but you can create an any() expressions like this

QPerson.person.attributes.any().attributeValue.eq("X")

This approach has the restriction that the join expression QPerson.person.attributes.any() can be used in only one filter. It has though the benefit that this expression is internally converted into a subquery which doesn't conflict with paging.

For multiple restrictions you will need to construct a subquery expression explicitly like this

QPersonAttribute attribute = QPersonAttribute.personAttribute;
new JPASubQuery().from(attribute)
    .where(attribute.in(person.attributes),
           attribute.attributeName().name.toLowerCase().eq("eye color"),
           attribute.attributeValue.toLowerCase().eq("blue"))
     .exists()

In addition to QueryDslPredicateExecutor you can also use Querydsl queries via Spring Data like this

public class CustomerRepositoryImpl
 extends QuerydslRepositorySupport
 implements CustomerRepositoryCustom {

    public Iterable<Customer> findAllLongtermCustomersWithBirthday() {
        QCustomer customer = QCustomer.customer;
        return from(customer)
           .where(hasBirthday().and(isLongTermCustomer()))
           .list(customer);
    }
}

Example taken from here https://blog.42.nl/articles/spring-data-jpa-with-querydsl-repositories-made-easy/