spring data jpa Specification 复杂查询+分页查询

当Repository接口继承了JpaSpecificationExecutor后，我们就可以使用如下接口进行分页查询：

    /**
     * Returns a {@link Page} of entities matching the given {@link Specification}.
     *
     * @param spec can be {@literal null}.
     * @param pageable must not be {@literal null}.
     * @return never {@literal null}.
     */
    Page<T> findAll(@Nullable Specification<T> spec, Pageable pageable);

结合jpa-spec可以很容易构造出Specification:

jpa-spec github地址:https://github.com/wenhao/jpa-spec

public Page<Person> findAll(SearchRequest request) {
    Specification<Person> specification = Specifications.<Person>and()
            .eq(StringUtils.isNotBlank(request.getName()), "name", request.getName())
            .gt(Objects.nonNull(request.getAge()), "age", 18)
            .between("birthday", new Date(), new Date())
            .like("nickName", "%og%", "%me")
            .build();

    return personRepository.findAll(specification, new PageRequest(0, 15));
}

单表查询确实很简单，但对复杂查询，就复杂上些了：

public List<Phone> findAll(SearchRequest request) {
    Specification<Phone> specification = Specifications.<Phone>and()
        .eq(StringUtils.isNotBlank(request.getBrand()), "brand", "HuaWei")
        .eq(StringUtils.isNotBlank(request.getPersonName()), "person.name", "Jack")
        .build();

    return phoneRepository.findAll(specification);
}

这里主表是phone，使用了person的name做条件,使用方法是person.name。

jpa-spec内部会分析person.name，如下代码：

public From getRoot(String property, Root<T> root) {
        if (property.contains(".")) {
            String joinProperty = StringUtils.split(property, ".")[0];
            return root.join(joinProperty, JoinType.LEFT);
        } else {
            return root;
        }
    }

就可看到它用了root.join，那就有一个问题，如果有两个person字段的条件，那就要再join一次，就会生成这样的sql:

select * from phone left outer join person on XX=XX left outer join person XX=XX.

这样肯定不满足需求。这应该也是jpa-spec的一个bug吧

为了解决这个问题，可以使用它提供的另一种方式查询：

public List<Phone> findAll(SearchRequest request) {
    Specification<Person> specification = Specifications.<Person>and()
        .between("age", 10, 35)
        .predicate(StringUtils.isNotBlank(jack.getName()), ((root, query, cb) -> {
            Join address = root.join("addresses", JoinType.LEFT);
            return cb.equal(address.get("street"), "Chengdu");
        }))
        .build();

    return phoneRepository.findAll(specification);
}

这要就可以解决大多数情况了，除了分页

看下正常的单表分页+排序查询：

public Page<Person> findAll(SearchRequest request) {
    Specification<Person> specification = Specifications.<Person>and()
            .eq(StringUtils.isNotBlank(request.getName()), "name", request.getName())
            .gt("age", 18)
            .between("birthday", new Date(), new Date())
            .like("nickName", "%og%")
            .build();

    Sort sort = Sorts.builder()
        .desc(StringUtils.isNotBlank(request.getName()), "name")
        .asc("birthday")
        .build();

    return personRepository.findAll(specification, new PageRequest(0, 15, sort));
}

如果在此基础上增加关联，如下代码：

public Page<Person> findAll(SearchRequest request) {
        Specification<Person> specification = Specifications.<Person>and()
                .predicate(StringUtils.isNotBlank(jack.getName()), ((root, query, cb) -> {
                    Join address = root.join("addresses", JoinType.LEFT);
                    return cb.equal(address.get("street"), "Chengdu");
                }))
                .eq(StringUtils.isNotBlank(request.getName()), "name", request.getName())
                .gt("age", 18)
                .between("birthday", new Date(), new Date())
                .like("nickName", "%og%")
                .build();

        Sort sort = Sorts.builder()
                .desc(StringUtils.isNotBlank(request.getName()), "name")
                .asc("birthday")
                .build();

        return personRepository.findAll(specification, new PageRequest(0, 15, sort));
    }

就会发现addresses的延迟加载失效，生成很多查询addresses的语句，解决方案如下：

public Page<Person> findAll(SearchRequest request) {
        Specification<Person> specification = Specifications.<Person>and()
                .predicate(StringUtils.isNotBlank(jack.getName()), ((root, query, cb) -> {
                    Join address;
                    if (Long.class != query.getResultType()) {
                        address = (Join) root.fetch("addresses", JoinType.LEFT);
                    } else {
                        address = root.join("addresses", JoinType.LEFT);
                    }
                    return cb.equal(address.get("street"), "Chengdu");
                }))
                .eq(StringUtils.isNotBlank(request.getName()), "name", request.getName())
                .gt("age", 18)
                .between("birthday", new Date(), new Date())
                .like("nickName", "%og%")
                .build();

        Sort sort = Sorts.builder()
                .desc(StringUtils.isNotBlank(request.getName()), "name")
                .asc("birthday")
                .build();

        return personRepository.findAll(specification, new PageRequest(0, 15, sort));
    }

至此，用Specification查询就应该够用了，再配合JpaRepository （SimpleJpaRepository）提供的方法 和@Query注解方法，和criteria api查询，这四种JPA查询就可以解决大多数应用问题了。