“不在具有JPA条件API的`subquery`”语句中

Question

我在Java EE应用程序中使用JPA和Criteria API查询数据库（PostgreSQL）。我实现了一个树作为一个闭包表，我试图获得根节点。下面是我的模式（省略了无用的字段）：

NeedsTreev2     :
id              | primary key

NeedNode        :
id              | primary key
needstree_id    | foreign key references needstreev2(id)

NeedLink        :
ancestor_id     | foreign key references neednode(id)
descendant_id   | foreign key references neednode(id)
needstree_id    | foreign key references needstreev2(id)

public class NeedsTreev2 {
    @Id
    private Long id;
}

public class NeedNode {
    @Id
    private Long id;
}

public class NeedLink {
    @ManyToOne
    private NeedNode ancestor;
    @ManyToOne
    private NeedNode descendant;
    @ManyToOne
    private NeedsTreev2;
}

SELECT nNode.* FROM neednode nNode
               INNER JOIN needstreev2 nTree
                       ON nNode.needstree_id = nTree.id

               WHERE nTree.id = ?
                 AND nNode.id NOT IN
                (SELECT nLink.descendant_id FROM needlink nLink
                                            WHERE nLink.ancestor_id != nLink.descendant_id)
               ;

public List<NeedNode> getRootsByTree(NeedsTreev2 tree) {
        List<NeedNode> ret;

        CriteriaBuilder cb = this.getEntityManager().getCriteriaBuilder();
        CriteriaQuery<NeedNode> cq = cb.createQuery(NeedNode.class);

        Root<NeedNode> nNode = cq.from(NeedNode.class);

        /* Here we define the subquery */
        Subquery<NeedNode> sq = cq.subquery(NeedNode.class);
        Root<NeedLink> nLink = sq.from(NeedLink.class);
        sq.where(cb.notEqual(nLink.get(NeedLink_.ancestor), nLink.get(NeedLink_.descendant)));
        sq.select(nLink.get(NeedLink_.descendant));
        /* End of subquery */

        Predicate[] p = {
            cb.equal(nNode.get(NeedNode_.needsTree), tree),
            cb.not(cb.in(nNode).value(sq)) /* This is where the problem occurs */
        };

        cq.where(cb.and(p));

        TypedQuery<NeedNode> query = this.getEntityManager().createQuery(cq);
        ret = query.getResultList();

        return (ret);
    }

org.eclipse.persistence.exceptions.QueryException
Exception Description: Illegal use of getField() [NEEDNODE.ID] in expression.
Query: ReadAllQuery(referenceClass=NeedNode )

Answer 1

我就是这样修复问题的：我主要选择了子查询中的neednode.id而不是完整对象。这样，in语句就可以工作了。

public List<NeedNode> getRootsByTree(NeedsTreev2 tree) {
    List<NeedNode> ret;

    CriteriaBuilder cb = this.getEntityManager().getCriteriaBuilder();
    CriteriaQuery<NeedNode> cq = cb.createQuery(NeedNode.class);

    Root<NeedNode> nNode = cq.from(NeedNode.class);

    Subquery<Long> sq = cq.subquery(Long.class);
    Root<NeedLink> nLink = sq.from(NeedLink.class);
    Join<NeedLink, NeedNode> d = nLink.join(NeedLink_.descendant, JoinType.INNER);
    sq.where(cb.notEqual(nLink.get(NeedLink_.ancestor), nLink.get(NeedLink_.descendant)));
    sq.select(d.get(NeedNode_.id));
    sq.distinct(true);

    Predicate[] p = {
        cb.equal(nNode.get(NeedNode_.needsTree), tree),
        nNode.get(NeedNode_.id).in(sq).not()
    };

    cq.where(cb.and(p));

    TypedQuery<NeedNode> query = this.getEntityManager().createQuery(cq);
    ret = query.getResultList();

    return (ret);
}

因此代码需要join变量和子查询来返回long而不是neednode。它是这样工作的，虽然我不明白为什么它不像问题中写的那样工作。IMO，使子查询选择ids会使条件查询的“类型安全”特性稍稍松动一些。