Spring Data JPA + QueryDSL 查询优化

Question

在使用与 Spring Data JPA 集成的 QueryDSL 时，我遇到了一个奇怪的行为：

我在 Project 和 Person 之间有一个 ManyToOne 关系。如果我通过所有者 ID（外键）获得属于用户的所有项目，一切都会按预期工作：

QProject project = QProject.project;
QPerson owner = project.owner;
List<Project> projects = from(project).leftJoin(owner).fetch()
    .where(owner.id.eq(id)).list(project);

生成的查询：

select
    project0_.id as id1_1_0_,
    person1_.id as id1_0_1_,
    project0_.creation_date as creation2_1_0_,
    project0_.name as name3_1_0_,
    project0_.owner as owner4_1_0_,
    person1_.name as name2_0_1_
from
    project project0_
left outer join
    person person1_
        on project0_.owner=person1_.id
where
    project0_.owner=?

但是，假设我们想通过一个不是外键的字段（例如所有者的姓名）来获取属于某个人的所有项目：

QProject project = QProject.project;
QPerson owner = project.owner; 
List<Project> projects = from(project).leftJoin(owner).fetch()
    .where(owner.name.eq(name)).list(project);

在这些情况下，表 Person 不必要地连接了两次（注意 person1_ 和 person2_）：

select
    project0_.id as id1_1_0_,
    person1_.id as id1_0_1_,
    project0_.creation_date as creation2_1_0_,
    project0_.name as name3_1_0_,
    project0_.owner as owner4_1_0_,
    person1_.name as name2_0_1_
from
    project project0_
left outer join
    person person1_
        on project0_.owner=person1_.id cross
join
    person person2_
where
    project0_.owner=person2_.id
    and person2_.name=?

知道为什么会发生这种情况以及如何避免吗？

Answer 1

您需要创建一个别名以确保在 where 部分重复使用第一个连接

QProject project = QProject.project;
QPerson owner = new QPerson("owner");
List<Project> projects = from(project)
    .leftJoin(project.owner, owner).fetch()
    .where(owner.name.eq(name))
    .list(project);