在一些E2E测试中，我遇到了一个问题。假设我有以下JPQL查询：
Query query = entityManager.createQuery(
" select d from Document d left join d.audit da " +
" where " +
" (cast(:startDate as java.time.ZonedDateTime)) is null " +
" or truncate_for_minutes(da.dateCreate, 'UTC') >= " +
" truncate_for_minutes(:startDate, 'UTC')")
.setParameter("startDate", ZonedDateTime.now());
在查询字符串中，我使用名为startDate的命名参数。上面的查询有效。但是，如果我通过null，则会引发以下异常：
null如果不进行类型转换，则会引发以下异常：
Caused by: org.postgresql.util.PSQLException:
ERROR: cannot cast type bytea to timestamp without time zone
如果不检查null，则会引发以下异常：
Caused by: org.postgresql.util.PSQLException:
ERROR: could not determine data type of parameter $1
我通过使用Caused by: org.postgresql.util.PSQLException:
ERROR: function pg_catalog.timezone(unknown, bytea) does not exist
No function matches the given name and argument types.
You might need to add explicit type casts.
字符串在Spring Data存储库中使用此查询。函数@Query-只是PostgreSQL函数truncate_for_minute(...)的一个小定制。
我知道我可以实现自定义存储库并动态构建查询字符串，但是为什么不能在JPQL字符串中检查date_trunc(...) null？也许有办法做到这一点？
我的环境：
答案 0 :(得分：0)
另一种解决方案是使用Criteria API并动态构建查询。
@Repository
@RequiredArgsConstructor
public class CustomDocumentRepositoryImpl implements CustomDocumentRegistry {
private final EntityManager em;
@Override
public Page<Document> findDocumentsForExpertByFilter(SearchDocumentCriteria criteria,
Pageable pageable) {
final String AUDIT_TABLE = "...";
final String USER_TABLE = "...";
final String ID_FIELD = "id";
final String FIRST_NAME_FIELD = "...";
final String LAST_NAME_FIELD = "...";
final String MIDDLE_NAME_FIELD = "...";
final String WORKSTATION_FIELD = "...";
final String DATE_CREATE_FIELD = "...";
final String LIKE_MASK = "%%%s%%";
CriteriaBuilder cb = em.getCriteriaBuilder();
CriteriaQuery<Document> query = cb.createQuery(Document.class);
Root<Document> root = query.from(Document.class);
Path<ZonedDateTime> dateCreatePath =
root.get(AUDIT_TABLE).get(DATE_CREATE_FIELD);
Path<String> lastNamePath =
root.get(AUDIT_TABLE).get(USER_TABLE).get(LAST_NAME_FIELD);
Path<String> firstNamePath =
root.get(AUDIT_TABLE).get(USER_TABLE).get(FIRST_NAME_FIELD);
Path<String> middleNamePath =
root.get(AUDIT_TABLE).get(USER_TABLE).get(MIDDLE_NAME_FIELD);
root.fetch(AUDIT_TABLE, JoinType.LEFT)
.fetch(USER_TABLE, JoinType.LEFT);
Predicate documentIdsPredicate;
List<Long> documentIds = criteria.getIds();
if (isNull(documentIds) || documentIds.isEmpty()) {
documentIdsPredicate = cb.isNotNull(root.get(ID_FIELD));
} else {
documentIdsPredicate = root.get(ID_FIELD).in(criteria.getIds());
}
Predicate startDatePredicate;
ZonedDateTime startDate = criteria.getStartDate();
if (isNull(startDate)) {
startDatePredicate = cb.isNotNull(dateCreatePath);
} else {
startDatePredicate = cb.greaterThanOrEqualTo(dateCreatePath, startDate);
}
Predicate endDatePredicate;
ZonedDateTime endDate = criteria.getEndDate();
if (isNull(endDate)) {
endDatePredicate = cb.isNotNull(dateCreatePath);
} else {
endDatePredicate = cb.lessThanOrEqualTo(dateCreatePath, endDate);
}
Predicate lastNamePredicate = cb.like(cb.upper(lastNamePath),
format(LIKE_MASK, criteria.getLastName().toUpperCase()));
Predicate firstNamePredicate = cb.like(cb.upper(firstNamePath),
format(LIKE_MASK, criteria.getFirstName().toUpperCase()));
Predicate middleNamePredicate = cb.like(cb.upper(middleNamePath),
format(LIKE_MASK, criteria.getMiddleName().toUpperCase()));
Predicate fullNamePredicate =
cb.and(lastNamePredicate, firstNamePredicate, middleNamePredicate);
Predicate compositePredicate = cb.and(
fullNamePredicate,
documentIdsPredicate,
startDatePredicate,
endDatePredicate
);
query.where(compositePredicate);
Query limitedQuery = em.createQuery(query
.orderBy(cb.desc(root.get(AUDIT_TABLE).get(DATE_CREATE_FIELD))))
.setFirstResult(nonNull(criteria.getSize()) ?
criteria.getPage() * criteria.getSize() :
criteria.getPage());
if (nonNull(criteria.getSize())) {
limitedQuery.setMaxResults(criteria.getSize());
}
List<Document> documents = limitedQuery.getResultList();
return new PageImpl<>(documents, pageable, criteria.getSize());
}
}
生成以下SQL：
select
document0_.id as id1_3_0_,
user1_.id as id1_13_1_,
document0_.created_dt as created_2_3_0_,
document0_.updated_dt as updated_3_3_0_,
document0_.created_user_id as created_6_3_0_,
document0_.updated_user_id as updated_7_3_0_,
document0_.name as name4_3_0_,
user1_.first_name as first_na2_13_1_,
user1_.last_name as last_nam3_13_1_,
user1_.middle_name as middle_n5_13_1_
from
some_scheme.document document0_
left outer join
some_scheme.s_user user1_
on document0_.created_user_id=user1_.id cross
join
some_scheme.s_user user2_
where
document0_.created_user_id=user2_.id
and (
upper(user2_.last_name) like '%LASTNAME%'
)
and (
upper(user2_.first_name) like '%FIRSTNAME%'
)
and (
upper(user2_.middle_name) like '%MIDDLENAME%'
)
and (
document0_.id in (
2 , 1
)
)
and document0_.created_dt>=...
and document0_.created_dt<=...
order by
document0_.created_dt desc limit 10;
答案 1 :(得分：0)
就我而言，问题如下。我注册了SQL函数date_trunc的自定义：
public class CustomSqlFunction implements MetadataBuilderContributor {
@Override
public void contribute(MetadataBuilder metadataBuilder) {
metadataBuilder.applySqlFunction(
"truncate_for_minutes",
new SQLFunctionTemplate(
StandardBasicTypes.TIMESTAMP,
"date_trunc('minute', (?1 AT TIME ZONE ?2))"
)
);
}
}
如果将StandardBasicTypes.TIMESTAMP更改为ZonedDateTime.INSTANCE并在一个参数中传递ZonedDateTime，则JPQL查询中的比较不会引起错误：
public class CustomSqlFunction implements MetadataBuilderContributor {
@Override
public void contribute(MetadataBuilder metadataBuilder) {
metadataBuilder.applySqlFunction(
"truncate_for_minutes",
new SQLFunctionTemplate(
ZonedDateTimeType.INSTANCE,
"date_trunc('minute', ?1)"
)
);
}
}