JPA多对多ManyToMany关联下保存Agency后无法获取Client实体ID

问题根因 #

1. 你使用agencyService.getOne()获取的是JPA懒加载代理对象，代理对象的集合操作可能导致持久化上下文的实例属性同步延迟。
2. 你直接持有手动new的Client实例取ID，级联保存虽然会填充持久化上下文中Client实例的ID，但受HashSet存储逻辑、代理对象实例隔离的影响，你持有的原始实例不会被同步赋值。
3. 额外隐患：你Client实体的equals和hashCode基于name字段生成，若存在重复name会导致Agency的clients集合中存入重复Client数据，也会影响实例匹配。

解决方案 #

方案1（保留级联保存逻辑，改动最小） #

修改业务代码如下：

@Override
@Transactional
public Long save(ClientCreateDto clientCreateDto) {
    try {
        Client client = new Client();
        client.setName(clientCreateDto.getName());
        Address address = new Address();
        address.setFirstLine(clientCreateDto.getAddress().getFirstLine());
        address.setTown(clientCreateDto.getAddress().getTown());
        address.setPostcode(clientCreateDto.getAddress().getPostcode());
        client.setAddress(address);
        client.setEmail(clientCreateDto.getEmail());
        client.setBillingEmail(clientCreateDto.getBillingEmail());
        client.setTelephone(clientCreateDto.getTelephone());
        client.setLogo(clientCreateDto.getLogo());
        client.setCreatedBy("System");
        client.setStatus(Status.ACTIVE);

        client.setService(service.getOne(clientCreateDto.getServiceId()));
        // 替换getOne为findById，获取真实实体对象而非代理
        Agency agency = agencyRepository.findById(clientCreateDto.getAgencyId())
                .orElseThrow(() -> new BusinessValidationException("机构不存在"));


        agency.addClient(client);
        agencyRepository.saveAndFlush(agency);
        // 从agency的clients集合中匹配获取已持久化的client实例取ID
        return agency.getClients().stream()
                .filter(c -> c.getName().equals(client.getName()))
                .findFirst()
                .orElseThrow(() -> new BusinessValidationException("客户保存失败"))
                .getId();
    } catch (Exception exception) {
        throw new BusinessValidationException(exception.getMessage());
    }
}

如果要更稳妥，可以在Agency的@ManyToMany注解的级联配置中追加CascadeType.REFRESH：

@ManyToMany(cascade = {
        CascadeType.PERSIST,
        CascadeType.MERGE,
        CascadeType.REFRESH
})
@JoinTable(name = "agency_client",
        joinColumns = @JoinColumn(name = "agency_id"),
        inverseJoinColumns = @JoinColumn(name = "client_id")
)
private Set<Client> clients = new HashSet<>();

方案2（性能更稳定，避免级联坑） #

不依赖级联保存Client，先单独持久化Client再关联：

@Override
@Transactional
public Long save(ClientCreateDto clientCreateDto) {
    try {
        Client client = new Client();
        client.setName(clientCreateDto.getName());
        Address address = new Address();
        address.setFirstLine(clientCreateDto.getAddress().getFirstLine());
        address.setTown(clientCreateDto.getAddress().getTown());
        address.setPostcode(clientCreateDto.getAddress().getPostcode());
        client.setAddress(address);
        client.setEmail(clientCreateDto.getEmail());
        client.setBillingEmail(clientCreateDto.getBillingEmail());
        client.setTelephone(clientCreateDto.getTelephone());
        client.setLogo(clientCreateDto.getLogo());
        client.setCreatedBy("System");
        client.setStatus(Status.ACTIVE);

        client.setService(service.getOne(clientCreateDto.getServiceId()));
        // 先保存client，直接拿到ID
        client = clientRepository.saveAndFlush(client);
        
        Agency agency = agencyRepository.findById(clientCreateDto.getAgencyId())
                .orElseThrow(() -> new BusinessValidationException("机构不存在"));
        agency.addClient(client);
        agencyRepository.saveAndFlush(agency);

        return client.getId();
    } catch (Exception exception) {
        throw new BusinessValidationException(exception.getMessage());
    }
}

内容的提问来源于stack exchange，提问作者Mapfunde Venon