我正在尝试让嵌套投影在Spring Boot中工作。我有2个实体,父母
和孩子
,而父母
与孩子
有一个单向的@OneTo多
关系。
以下是类:(使用Lombok注释)
@Entity
@Data @NoArgsConstructor
public class Parent {
@Id
@GeneratedValue
private long id;
private String basic;
private String detail;
@OneToMany(fetch = FetchType.EAGER)
private List<Child> children;
public Parent(String basic, String detail, List<Child> children) {
this.basic = basic;
this.detail = detail;
this.children = children;
}
}
@Entity
@Data @NoArgsConstructor
public class Child {
@Id
@GeneratedValue(strategy = GenerationType.TABLE)
private long id;
private String basic;
private String detail;
public Child(String basic, String detail) {
this.basic = basic;
this.detail = detail;
}
}
当我在不投影的情况下获取数据时,我得到以下内容:
[
{
"id": 1,
"basic": "parent-basic-1",
"detail": "parent-detail-1",
"children": [
{
"id": 1,
"basic": "child-basic-1",
"detail": "child-detail-1"
},
{
"id": 2,
"basic": "child-basic-2",
"detail": "child-detail-2"
}
]
},
{
"id": 2,
"basic": "parent-basic-2",
"detail": "parent-detail-2",
"children": [
{
"id": 3,
"basic": "child-basic-3",
"detail": "child-detail-3"
},
{
"id": 4,
"basic": "child-basic-4",
"detail": "child-detail-4"
}
]
}
目标如下:
{
"id": 1,
"basic": "parent-basic-1",
"children": [1,2]
},
{
"id": 2,
"basic": "parent-basic-2",
"children": [3,4]
}
然而,似乎完全不可能实现这一目标。
@Value
public class ParentDto {
long id;
String basic;
// wanted to get it to work with just Child instead of ChildDto first, before getting ChildDto to work
Collection<Child> children;
public ParentDto(long id, String basic, Collection<Child> children) {
this.id = id;
this.basic = basic;
this.children = children;
}
}
// Constructor Projection in Repository
@Query("select new whz.springbootdemo.application.constructor_projection.ParentDto(p.id, p.basic, p.children) from Parent p")
List<ParentDto> findAllConstructorProjected();
但这会导致以下错误:
could not prepare statement; SQL [select parent0_.id as col_0_0_, parent0_.basic as col_1_0_, . as col_2_0_ from parent parent0_ inner join parent_children children1_ on parent0_.id=children1_.parent_id inner join child child2_ on children1_.children_id=child2_.id]; nested exception is org.hibernate.exception.SQLGrammarException: could not prepare statement
// Dynamic Projection in Repository
List<ParentDto> findAllDynamicProjectionBy();
导致以下错误:
org.hibernate.hql.internal.ast.QuerySyntaxException:
Unable to locate appropriate constructor on class [whz.springbootdemo.application.constructor_projection.ParentDto].
Expected arguments are: <b>long, java.lang.String, whz.springbootdemo.application.child.Child</b>
[select new whz.springbootdemo.application.constructor_projection.ParentDto(generatedAlias0.id, generatedAlias0.basic, children) from whz.springbootdemo.application.parent.Parent as generatedAlias0 left join generatedAlias0.children as children]; nested exception is java.lang.IllegalArgumentException: org.hibernate.hql.internal.ast.QuerySyntaxException: Unable to locate appropriate constructor on class [whz.springbootdemo.application.constructor_projection.ParentDto]. Expected arguments are: long, java.lang.String, whz.springbootdemo.application.child.Child [select new whz.springbootdemo.application.constructor_projection.ParentDto(generatedAlias0.id, generatedAlias0.basic, children) from whz.springbootdemo.application.parent.Parent as generatedAlias0 left join generatedAlias0.children as children]
这基本上告诉我执行了一个连接,但是值没有按父节点的id分组,因此产生了x行,其中x是父节点拥有的子节点数,每个子节点都有父节点的基本信息和一个子节点信息。
// Interface Projection in Repository
List<ParentDtoInterface> findAllInterfaceProjectedBy();
public interface ParentDtoInterface {
long getId();
String getBasic();
List<ChildDtoInterface> getChildren();
}
public interface ChildDtoInterface {
long getId();
}
它导致:
[
{
"id": 1,
"children": [
{
"id": 1
},
{
"id": 2
}
],
"basic": "parent-basic-1"
},
{
"id": 2,
"children": [
{
"id": 3
},
{
"id": 4
}
],
"basic": "parent-basic-2"
}
]
现在我对接口投影的问题是,它不仅会加载预期的属性,还会加载所有属性,但jackson只会序列化接口提供的那些属性,因为它使用类/接口定义。
父加载:(sql日志;参见第4行,加载了详细信息)
select
parent0_.id as id1_1_,
parent0_.basic as basic2_1_,
parent0_.detail as detail3_1_
from
parent parent0_
此外,接口投影似乎真的很慢(参见这个Stackoverflow问题),我仍然必须解压缩子级,因为它们被给出为[{id: 1},{id:2}],但我真的需要[1,2]。我知道我可以用@JsonIdtyApplication(alwaysAsId=true)
来做到这一点,但这只是一种解决方法。
我也有点困惑为什么数据在n个查询中加载-1个用于父母,另一个n(其中n是父母的数量)用于每个父母的孩子:
select
parent0_.id as id1_1_,
parent0_.basic as basic2_1_,
parent0_.detail as detail3_1_
from
parent parent0_
select
children0_.parent_id as parent_i1_2_0_,
children0_.children_id as children2_2_0_,
child1_.id as id1_0_1_,
child1_.basic as basic2_0_1_,
child1_.detail as detail3_0_1_
from
parent_children children0_
inner join
child child1_
on children0_.children_id=child1_.id
where
children0_.parent_id=?
//... omitting further child queries
我尝试过@OneTo很多(fetch=FetchType. LAZY)
和@Fetch(FetchType.JOINED)
-两者都给出了与上述相同的结果。
所以主要的问题是:有什么方法可以实现嵌套实体的Spring Boot投影,以便在尽可能少的查询中只加载所需的数据,在最好的情况下,我可以调整它,而不是加载List子项,我可以只加载List子项(也许通过Jpa查询,将连接的行按括号分组,让我们从子项中提取所需的数据?)。
我使用Hibernate和内存数据库。
感谢您的任何回答或提示!
编辑:澄清一下:我并没有试图找到一种方法来序列化所需格式的数据——这我已经可以实现了。主要关注点是只从数据库中加载必要的信息。
这将永远吸引孩子,但可以给你你想要的结果。
public interface SimpleParentProjection {
String getBasic();
String getDetail();
@Value("#{T(SimpleParentProjection).toId(target.getChildren())}")
String[] getChildren();
static String[] toId(Set<Child> childSet) {
return childSet.stream().map(c -> String.valueOf(c.getId())).toArray(String[]::new);
}
}