Spring冬眠 为什么多对多关系没有持续存在？

Question

我有一个实体 E1 定义为

@Entity
public class E1 {
    @Id
    @GeneratedValue
    public long id;
    @ManyToMany(cascade={CascadeType.PERSIST, CascadeType.MERGE})
    @JoinTable(
            name="e1_e2",
            joinColumns = @JoinColumn(name = "e2_id"),
            inverseJoinColumns = @JoinColumn(name = "e1_id")
    )
    public Set<E2> e2s = new HashSet<>();
}

实体 E2 定义为

@Entity
public class E2 {
    @Id
    @GeneratedValue
    public long id;
    @ManyToMany(mappedBy = "e2s")
    public Set<E1> e1s = new HashSet<>();
}

控制器定义为：

@RestController
@RequestMapping("/")
public class C1 {
    private final E1Repository e1Repository;
    private final E2Repository e2Repository;

    @PersistenceContext
    EntityManager em;

    @Autowired
    public C1(E1Repository e1Repository, E2Repository e2Repository) {
        this.e1Repository = e1Repository;
        this.e2Repository = e2Repository;
    }

    @Transactional
    @RequestMapping(method = POST)
    public void c(){
        E1 e1 = new E1();
        E2 e2 = new E2();

        e1Repository.save(e1);
        e2Repository.save(e2);

        em.refresh(e1);
        em.refresh(e2);

        e1.e2s.add(e2);
        e2.e1s.add(e1);

        e1Repository.save(e1);
        e2Repository.save(e2);

        em.refresh(e1);
        em.refresh(e2);
    }

}

（E1Repository和E2Repository，是@repositoyannotated接口，扩展JpaRepository并具有空体）。

当我在调试器中单步执行c方法时，我看到在最后两行em.refresh行之后，e1和e2都清除了它们的集合。

基于我在Stack Overflow上发现的其他问题，我尝试将E2定义为

@Entity
public class E2 {
    @Id
    @GeneratedValue
    public long id;
    @ManyToMany(cascade = CascadeType.PERSIST)
    @JoinTable(
            name="e1_e2",
            inverseJoinColumns = @JoinColumn(name = "e2_id"),
            joinColumns = @JoinColumn(name = "e1_id")
    )
    public Set<E1> e1s = new HashSet<>();
}

但这于事无补。

原问题(在我尝试测试上面的简化案例之前)我有一个类< code>Robot，定义如下:

@Entity
class Robot{
    @Id
    @GeneratedValue
    private long id;

    @ManyToMany(mappedBy="robots")
    private Set<Match> matches = new HashSet<>();
}

和一个类< code>Match类似于

@Entity
class Robot{
    @Id
    @GeneratedValue
    private long id;
    @ManyToMany(cascade={CascadeType.Persist,CascadeType.Merge})
    @JoinTable(
        name = "match_robot",
        joinColumns = {@JoinColumn(name = "Match_id")},
        inverseJoinColumns = {@JoinColumn(name = "Robot_id")}
)
private Set<Robot> robots = new HashSet<>();

类＜code＞结果＜code＞的定义类似于

@Entity
public class Result {
    @Id
    @GeneratedValue
    private long id;

    @ManyToOne(optional = false)
    private Match match;

    @ManyToOne(optional = false)
    @NotNull(groups = {Default.class, Creating.class})
    private Robot robot;
}

并尝试保存关系，如下所示：

resultRepository.save(result);
match.getRobots().add(result.getRobot());
result.getRobot().getMatches().add(match);
robotRepository.save(result.getRobot());
matchRepository.save(match);
entityManager.refresh(result);
entityManager.refresh(result.getRobot());
entityManager.refresh(match);

其中< code>*Repository是spring创建的实现JpaRepository的beans。运行这段代码时，hibernate会为< code>result对象运行一个insert语句(hibernate会将其所有sql命令输出到控制台)，但不会为“match_robot”表运行任何语句。在这段代码之前，< code>result处于hibernate实体生命周期的瞬态，< code>match处于持久状态，< code>robot处于持久状态，< code>result的< code>match和< code>robot属性已分别设置为< code>match和< code>robot。

基于栈溢出和其他站点的其他问题，我还尝试将matches变量定义为

@ManyToMany(cascade = CascadeType.PERSIST)
@JoinTable(
        name = "match_robot",
        inverseJoinColumns = {@JoinColumn(name = "Match_id")},
        joinColumns = {@JoinColumn(name = "Robot_id")}
)
private Set<Match> matches = new HashSet<>();

除此之外，我看到的唯一建议是在坚持之前确保多对多关系中的两个实体彼此一致，但是，据我所知，我正在这样做。

我该如何坚持这段感情？

编辑：上下文的完整方法：

@Transactional
@RequestMapping(value = "/{match:[0-9]+}/results", method = RequestMethod.POST)
public ResultResource createResult(@PathVariable Match match,
                                   @Validated(Result.Creating.class)
                                   @RequestBody Result result) {
    if (match == null) throw new ResourceNotFoundException();

    result.setScorecard(scorecardRepository
            .findById(result.getScorecard().getId()));
    if (result.getScorecard() == null) {
        throw new ScorecardDoesNotExistException();
    }

    result.setMatch(match);

    //remove null scores
    result.getScores().removeIf(
            fieldResult -> fieldResult.getScore() == null
    );

    //replace transient robot with entity from database
    Robot existingRobot = robotRepository
            .findByNumberAndGame(result.getRobot().getNumber(),result.getRobot().getGame());
     if (existingRobot == null) { //create new robot
        //find team for this robot
        Team existingTeam = teamRepository
                .findByNumberAndGameType(
                        result.getRobot().getNumber(),
                        result.getScorecard().getGame().getType());
        if (existingTeam == null) {
            Team team = new Team();
            team.setNumber(result.getRobot().getNumber());
            team.setGameType(result.getMatch().getEvent().getGame().getType());
            team.setDistrict(result.getMatch().getEvent().getDistrict());
            teamRepository.save(team);
            result.getRobot().setTeam(team);
        }
    else result.getRobot().setTeam(existingTeam);
             result.getRobot().setGame(result.getMatch().getEvent().getGame());

        robotRepository.save(result.getRobot());
        entityManager.refresh(result.getRobot());
    } else result.setRobot(existingRobot);
    List<Robot> all = robotRepository.findAll();

    //replace transient FieldSections with entities from database
    //todo: reduce database hits
    //noinspection ResultOfMethodCallIgnored
    result.getScores().stream()
          .peek(fieldResult -> fieldResult.setField(
                  fieldSectionRepository.findByIdAndScorecard(
                          fieldResult.getField().getId(),
                          result.getScorecard())))
              .peek(fieldResult -> {
              if (fieldResult.getField() == null)
                  throw new ScoresDoNotExistException();
          })
          .forEach(fieldResult->fieldResult.setResult(result));


    if (!result.scoresMatchScorecardSections()) {
        throw new ScoresDoNotMatchScorecardException();
    }

    if (!result.allMissingScoresAreOptional()) {
        throw new RequiredScoresAbsentException();
    }

    if (!result.gameMatchesScorecard()) {
        throw new GameDoesNotMatchScorecardException();
    }

    resultRepository.save(result);
    match.getRobots().add(result.getRobot());
    result.getRobot().getMatches().add(match);
    robotRepository.save(result.getRobot());
    matchRepository.save(match);
    entityManager.refresh(result);
    entityManager.refresh(result.getRobot());
    entityManager.refresh(match);
    return new ResultResourceAssembler().toResource(result);
}

Answer 1

JpaRepository.save（） 不会将更改刷新到数据库。有必要使用 JpaRepository.flush（） 或 JpaRepository.saveAndFlush（） 冲洗更改。