Spring Data JPA坑点记录
一、分页查询
场景:动态查询,分页查询,根据传入不同的状态,分别查询不同数据表,并且在传入page对象之前用map进行VO转换。而pageable的使用地方不同影响到了分页数据的正确性,以此进行探讨。
- pageable使用于new PageImpl<>中,且直到最后才将List -> Page
- pageable使用于findAll()中
前提:Page对象封于VO内,返回数据包括了分页数据
@ApiModelProperty("记录")
private Page<ActivityRecordVO> activityRecordVOList;
@ApiModelProperty("数量")
private Integer num = 0;
@ApiModelProperty("金额")
private BigDecimal totalMoney = BigDecimal.valueOf(0);
错误运用:
List<ActivityRecordVO> activityRecordVOList = new ArrayList<>();
if (receiveSendRecordRequestVO.getSendOrReceiveType() == SendOrReceiveType.RECEIVE) {
List<ChallengeRecord> challengeRecordList = challengeRecordDao.findByUserIdAndDeleteType(userId,
DeleteType.FALSE);
if (!CollectionUtils.isEmpty(challengeRecordList)) {
activityRecordVOList = challengeRecordList.stream()
.map(this::challengeRecordToActivityRecordVO)
.collect(Collectors.toList());
}
} else if (receiveSendRecordRequestVO.getSendOrReceiveType() == SendOrReceiveType.SEND) {
List<Activity> activityList = activityDao.findByUserIdAndDeleteType(userId, DeleteType.FALSE);
if (!CollectionUtils.isEmpty(activityList)) {
activityRecordVOList = activityList.stream()
.map(this::activityTOActivityRecordVO)
.collect(Collectors.toList());
}
}
activityReceiveSendRecordVO.setActivityRecordVOList(new PageImpl<>(activityRecordVOList,
pageable, activityRecordVOList.size()));
解析:传入的pageable只在set进VO的时候,用new PageIml将List转为page对象,前端报的问题 虽然总页数、总条数均为正确,但第一页的条数是全部 ,数据异常!
正确参考做法:
采用Specifications先根据查询条件动态查询并map出相应分页对象(此块代码因需求而异),这时 findAll 传入的pageable是生效的,便会显现正确的分页信息。
代码块参考:
xxxCommonSpecUtil 是自封的specification工具类,与原生spring data jpa原生查询方法类似。
Page<ActivityRecordVO> page = new PageImpl<>(activityRecordVOList, pageable, activityRecordVOList.size());
if (receiveSendRecordRequestVO.getSendOrReceiveType() == SendOrReceiveType.RECEIVE) {
Specifications<ChallengeRecord> spec = Specifications.where(
challengeCommonSpecUtil.equal("userId", userId))
.and(challengeCommonSpecUtil.equal("deleteType", DeleteType.FALSE));
page = challengeRecordDao.findAll(spec, pageable).map(this::challengeRecordToActivityRecordVO);
} else if (receiveSendRecordRequestVO.getSendOrReceiveType() == SendOrReceiveType.SEND) {
Specifications<Activity> spec = Specifications.where(
activityCommonSpecUtil.equal("userId", userId))
.and(activityCommonSpecUtil.equal("deleteType", DeleteType.FALSE));
page = activityDao.findAll(spec, pageable).map(this::activityTOActivityRecordVO);
}
注:activityReceiveSendRecordVO为封装的VO,包含了返回的Page对象
activityReceiveSendRecordVO.setActivityRecordVOList(page);
二、事务,更新实体,并查询
场景:@Transactional事务,更新用户余额(处理并发问题),更新完毕返回VO带上用户剩余余额,但却非更新后的余额。
更新逻辑代码:
userWebService.updateBalanceAfterTransaction(userId, transactionRecordAddVO.getMoney(),user.getBalance());
@Transactional
public void updateBalanceAfterTransaction(Integer userId, BigDecimal money, BigDecimal userBalance) {
int i = userDao.updateBalanceAfterTransaction(userId, money, userBalance);
if (i == 0) {
throw new ValidationException(MessageCodes.TRANSACTION_RECEIVE_IS_ERROR);
}
}
@Modifying
@Query(value = "update User u set u.balance = u.balance - ?2 where u.id = ?1 and u.balance = ?3")
int updateBalanceAfterTransaction(Integer userId, BigDecimal money, BigDecimal userBalance);
解析:
在jpa中使用 @Modifying ,虽然事务已经能够更新,但是在循环更新的时候,执行modify语句后的查询的实体仍然是没有更新的。
执行完modifying query, EntityManager可能会包含过时的数据,因为EntityManager不会自动清除实体。
只有添加clearAutomatically属性,EntityManager才会自动清除实体对象。
添加代码:
@Modifying(clearAutomatically = true)
改正后的示例代码:
@Modifying(clearAutomatically = true)
@Query(value = "update User u set u.balance = u.balance - ?2 where u.id = ?1 and u.balance = ?3")
int updateBalanceAfterTransaction(Integer userId, BigDecimal money, BigDecimal userBalance);
总结
使用了这么长时间spring data jpa,觉得Specifications巨好用,也不容易出错,也是我喜欢的编码风格,而new PageImpl<>()这种简单粗暴的方法我一般都用在查询数据关联太多表的情况,在最后直接返回,更深层次的还需要再探讨!