仅具有选定子项的 JPA 查询父项

Question

我有这些课程Parent 和Child

Parent.java

@Entity
@Table(name = "parents")
public class Parent implements Serializable {

private static final long serialVersionUID = 873222010704108510L;

@Id
@GeneratedValue(strategy = GenerationType.AUTO)
private Integer id;

private String name;

@OneToMany(cascade=CascadeType.ALL)
@JsonManagedReference
private Set<Child> childs;

// getters and setters

Child.java

@Entity
@Table(name = "childs")
public class Child implements Serializable {

private static final long serialVersionUID = -6756632049453970501L;

@Id
@GeneratedValue(strategy = GenerationType.AUTO)
private Integer id;

private String name;

@ManyToOne(fetch = FetchType.LAZY)
@JsonBackReference
private Parent parent;

// getters and setters

现在我的 SQL DB 有 2 个子级，child1 和 child2，它们都有相同的父级。

现在我想得到父母和唯一的孩子1，所以在我的ParentRepository.java 我有以下内容

@Query("select p from Parent p LEFT JOIN p.childs c where c.name = 'child1'")
public List<Parent> getParent();

但是当我调用它时，我得到了两个孩子，这是因为我看到 hibernate 正在运行 2 个查询，如下所示：

Query 1（我想要的查询）

/* select
    p 
from
    Parent p 
LEFT JOIN
    p.childs c 
where
    c.name = 'child1' */ select
        parent0_.id as id1_1_,
        parent0_.name as name2_1_ 
    from
        parents parent0_ 
    left outer join
        parents_childs childs1_ 
            on parent0_.id=childs1_.parent_id 
    left outer join
        childs child2_ 
            on childs1_.childs_id=child2_.id 
    where
        child2_.name='child1'

然后它运行这个查询 - 我不想要它，因为它正在获取所有的孩子

 select
    childs0_.parent_id as parent_i1_2_0_,
    childs0_.childs_id as childs_i2_2_0_,
    child1_.id as id1_0_1_,
    child1_.name as name2_0_1_,
    child1_.parent_id as parent_i3_0_1_ 
from
    parents_childs childs0_ 
inner join
    childs child1_ 
        on childs0_.childs_id=child1_.id 
where
    childs0_.parent_id=?

我怎样才能得到它，所以第二个查询没有运行，我得到了父母，只有我想要返回的孩子，即使没有匹配的孩子，我也想返回父母，这就是为什么我有左外加入

Answer 1

选项 1

换个角度想，在Child 上搜索并获得他们的Parent 怎么样。

将新的构造函数添加到Parent 类中

public Parent(Integer id, String name, Child child) {
    this.id = id;
    this.name = name;
    this.childs = new HashSet<>();
    this.childs.add(child);
}

public Parent() {
}

使用以下脚本更改 JPA 查询。

@Query("select new Parent(c.parent.id, c.parent.name, c) from Child c where c.name = 'child1'")
List<Parent> getParent();

Hibernate 为两个 Child 记录生成 3 个 SQL（所有 Child 计数中的多个查询）

Hibernate: select child0_.parent_id as col_0_0_, parent1_.name as col_1_0_, child0_.id as col_2_0_ from childs child0_ cross join parents parent1_ where child0_.parent_id=parent1_.id and child0_.name='child1'
Hibernate: select child0_.id as id1_0_0_, child0_.name as name2_0_0_, child0_.parent_id as parent_i3_0_0_ from childs child0_ where child0_.id=?
Hibernate: select child0_.id as id1_0_0_, child0_.name as name2_0_0_, child0_.parent_id as parent_i3_0_0_ from childs child0_ where child0_.id=?

如果只执行一条 SQL，可以升级下面的Parent 类和 Jpa 查询。

父构造函数：

public Parent(Integer id, String name, Integer childId, String childName) {
    this.id = id;
    this.name = name;
    this.childs = new HashSet<>();
    Child child = new Child();
    child.setId(childId);
    child.setName(childName);
    this.childs.add(child);
}

在 jpa 查询中添加了新参数

@Query("select new Parent(c.parent.id, c.parent.name, c.id, c.name) from Child c where c.name = 'child1'")
List<Parent> getParent();

Hibernate 生成单个查询。

Hibernate: select child0_.parent_id as col_0_0_, parent1_.name as col_1_0_, child0_.id as col_2_0_, child0_.name as col_3_0_ from childs child0_ cross join parents parent1_ where child0_.parent_id=parent1_.id and child0_.name='child1'

测试

@SpringBootTest
class DemoApplicationTests {

    @Autowired
    private ParentRepository parentRepository;
    
    @Test
    void contextLoads() {
        Parent parent = new Parent();
        parent.setName("p1");
        parent.setChilds(new HashSet<>());
        parent.getChilds().add(new Child("child1", parent));
        parent.getChilds().add(new Child("child2", parent));

        parentRepository.save(parent);

        parent = new Parent();
        parent.setName("p2");
        parent.setChilds(new HashSet<>());
        parent.getChilds().add(new Child("child1", parent));

        parentRepository.save(parent);

        List<Parent> parents = parentRepository.getParent();
        Assertions.assertEquals(2, parents.size());

        for (Parent parent1 : parents) {
            System.out.println(parent1);
        }
    }
}

输出

Parent(id=1, name=p1, childs=[Child(id=3, name=child1)])
Parent(id=4, name=p2, childs=[Child(id=5, name=child1)])

选项 2

获取所有子节点并在 java 端管理 Parent。

public interface ChildRepository extends JpaRepository<Child, Integer> {
    @Query("select c from Child c left join fetch c.parent where c.name = 'child1'")
    List<Child> search();
}

Hibernate 生成单个 SQL。

Hibernate: select child0_.id as id1_0_0_, parent1_.id as id1_1_1_, child0_.name as name2_0_0_, child0_.parent_id as parent_i3_0_0_, parent1_.name as name2_1_1_ from childs child0_ left outer join parents parent1_ on child0_.parent_id=parent1_.id where child0_.name='child1'

测试

@SpringBootTest
class DemoApplicationTests2 {

    @Autowired
    private ParentRepository parentRepository;

    @Autowired
    private ChildRepository childRepository;

    @Test
    void contextLoads() {
        Parent parent = new Parent();
        parent.setName("p1");
        parent.setChilds(new HashSet<>());
        parent.getChilds().add(new Child("child1", parent));
        parent.getChilds().add(new Child("child2", parent));

        parentRepository.save(parent);

        parent = new Parent();
        parent.setName("p2");
        parent.setChilds(new HashSet<>());
        parent.getChilds().add(new Child("child1", parent));

        parentRepository.save(parent);

        // search children
        List<Child> children = childRepository.search();

        // create parent hash map
        Map<Integer, Parent> parentMap = children.stream()
                .map(Child::getParent)
                .distinct()
                .peek(p -> {
                    p.setChilds(new HashSet<>());
                })
                .collect(Collectors.toMap(Parent::getId, p -> p));

        // add children manually
        for (Child child : children) {
            parentMap.get(child.getParent().getId()).getChilds().add(child);
        }

        // get parents
        Collection<Parent> parents = parentMap.values();

        Assertions.assertEquals(2, parents.size());

        for (Parent parent1 : parents) {
            System.out.println(parent1);
        }
    }
}

输出

Parent(id=1, name=p1, childs=[Child(id=3, name=child1)])
Parent(id=4, name=p2, childs=[Child(id=5, name=child1)])

额外

获取所有Parent 并填充匹配的Child 记录。我们需要对预定义的 Parent 构造函数进行一些更改，并将 where 的新 JPA 查询替换为 and 关键字。

父构造函数： 有一个额外的childId 控件来避免插入空子记录。

public Parent(Integer id, String name, Integer childId, String childName) {
    this.id = id;
    this.name = name;
    this.childs = new HashSet<>();
    if (childId != null) {
        Child child = new Child();
        child.setId(childId);
        child.setName(childName);
        this.childs.add(child);
    }
}

ParentRepository：没有where键，因为它限制了我们的结果，我们只需要and而不是它。

@Query("select new Parent(p.id, p.name, c.id, c.name) from Parent p left outer join Child c on c.parent = p and c.name = 'child2'")
List<Parent> getParent();

休眠查询： 刚刚执行了一个查询。

Hibernate: select parent0_.id as col_0_0_, parent0_.name as col_1_0_, child1_.id as col_2_0_, child1_.name as col_3_0_ from parents parent0_ left outer join childs child1_ on (child1_.parent_id=parent0_.id and child1_.name='child2')

测试

List<Parent> parents = parentRepository.getParent();
Assertions.assertEquals(2, parents.size());

for (Parent parent1 : parents) {
    System.out.println(parent1);
}

输出：我们正在搜索child2，一个记录有这个，另一个是空的。

Parent(id=1, name=p1, childs=[Child(id=2, name=child2)])
Parent(id=4, name=p2, childs=[])