Optimizing Querydsl BooleanExpression queries

Question

I've created a class with static methods that return Querydsl BooleanExpressions so I can simply pass these expressions to the findAll() or findOne() methods in my Spring Data repositories, as well as reuse them throughout my application. Like here: http://www.petrikainulainen.net/programming/spring-framework/spring-data-jpa-tutorial-part-five-querydsl/.

The issue that I'm having is that the BooleanExpression that's being returned from my methods is generating queries that seem to be sub-optimal (at least for my case).

Method that returns the BooleanExpression:

public static BooleanExpression byExternalIdAndProviderId(
     final String externalId, final Long providerId) {

  QEListing listing = QEListing.eListing;

  return listing.externalIds.any().externalId.equalsIgnoreCase(externalId)
        .and(listing.externalIds.any().provider.id.eq(providerId));
}

Query that is generated:

select elisting0_.id as id1_24_, 
    elisting0_.address1 as address2_24_, 
    elisting0_.address2 as address3_24_, 
    elisting0_.business_name as business4_24_, 
    elisting0_.city_id as city10_24_, 
    elisting0_.created_date as created5_24_, 
    elisting0_.latitude as latitude6_24_, 
    elisting0_.longitude as longitud7_24_, 
    elisting0_.master_ext_id as master11_24_, 
    elisting0_.modified_date as modified8_24_, 
    elisting0_.state_id as state12_24_, 
    elisting0_.zip as zip13_24_, 
    elisting0_.zip_4 as zip9_24_, 
    elisting0_1_.vanity_name as vanity1_34_, 
    elisting0_2_.account_id as account1_2_ 
from listings elisting0_ 
    left outer join vanity_names elisting0_1_ on elisting0_.id=elisting0_1_.listing_id 
    left outer join accounts_to_listings elisting0_2_ on elisting0_.id=elisting0_2_.listing_id 
where (
    exists (
        select 1 
        from external_ids eexternali1_ 
        where (
            eexternali1_.id in (
                select externalid2_.external_id 
                from listing_to_external_id externalid2_ 
                where elisting0_.id=externalid2_.listing_id)) 
                and lower(eexternali1_.external_id)='123456'
        )
    ) 
and (
    exists (
        select 1 
        from external_ids eexternali3_ 
        where (
            eexternali3_.id in (
                select externalid4_.external_id 
                from listing_to_external_id externalid4_ 
                where elisting0_.id=externalid4_.listing_id
            )
        ) 
        and eexternali3_.provider_id=1
    )
)

As opposed to something like this:

select elisting0_.id as id1_24_, 
    elisting0_.address1 as address2_24_, 
    elisting0_.address2 as address3_24_, 
    elisting0_.business_name as business4_24_, 
    elisting0_.city_id as city10_24_, 
    elisting0_.created_date as created5_24_, 
    elisting0_.latitude as latitude6_24_, 
    elisting0_.longitude as longitud7_24_, 
    elisting0_.master_ext_id as master11_24_, 
    elisting0_.modified_date as modified8_24_, 
    elisting0_.state_id as state12_24_, 
    elisting0_.zip as zip13_24_, 
    elisting0_.zip_4 as zip9_24_, 
    elisting0_1_.vanity_name as vanity1_34_, 
    elisting0_2_.account_id as account1_2_ 
from listings elisting0_ 
    left outer join vanity_names elisting0_1_ on elisting0_.id=elisting0_1_.listing_id 
    left outer join accounts_to_listings elisting0_2_ on elisting0_.id=elisting0_2_.listing_id 
    join listing_to_external_id a on elisting0_.id = a.listing_id
    join external_ids b on a.external_id = b.id
where lower(b.external_id) = '123456'
and b.provider_id = 1

Is there any way to generate a more optimal query, while still offering the re-usability of BooleanExpressions, and also using Spring Data's repositories?

Timo Westkämper Timo Westkämper · Accepted Answer · 2013-05-31T06:07:33

There might be a more optimal way, but joins can't be used, since they modify the result set. So if you find a way to optimize the any() subqueries in Querydsl, then please create a ticket in the GitHub issue tracker.

When the Querydsl usage is restricted to predicates, then there is no control over the joins.

Optimizing Querydsl BooleanExpression queries

1 Answers