MySQL和MariaDB中使用GROUP BY和ORDER BY的不同ResultSet

我有一张桌子：voj_submission具有以下结构：

CREATE TABLE IF NOT EXISTS `voj_submissions` (
    `submission_id` bigint(20) NOT NULL,
    `problem_id` bigint(20) NOT NULL,
    `uid` bigint(20) NOT NULL,
    `submission_judge_result` varchar(8) DEFAULT NULL,
) ENGINE=InnoDB AUTO_INCREMENT=1004 DEFAULT CHARSET=utf8;

并且表中的数据可能像这样：

SubmissionID ProblemID UserID JudgeResult
1000         1000      User1   AC
1001         1000      User1   WA
1002         1000      User2   CE
1003         1001      User1   AC

以下SQL希望获得问题的最新判断结果：

SELECT submission_id, problem_id, submission_judge_result 
FROM (
    SELECT * FROM 
    voj_submissions 
    ORDER BY submission_id DESC 
) s 
WHERE uid = 'User1'
AND problem_id >= 1000
AND problem_id < 1010
GROUP BY problem_id

在MySQL 5.6中,它可以正常工作并返回以下结果集：

SubmissionID ProblemID JudgeResult
1001         1000      WA
1003         1001      AC

但是在MariaDB 10.0.14中,它返回：

SubmissionID ProblemID JudgeResult
1000         1000      AC
1003         1001      AC

这意味着ORDER BY DESC不在MariaDB中执行.
我怎么解决这个问题？

参考：

> https://mariadb.com/kb/en/mariadb/group-by-trick-has-been-optimized-away/
> MySQL order by before group by

解决方法:

就像@bluefeet在此post中所说的：

Using an ORDER BY in a subquery is not the best solution to this problem.

The best solution to get the max(post_date) by author is to use a subquery to return the max date and then join that to your table on both the post_author and the max date.

因此解决方案应该是：

SELECT submission_id, s1.problem_id, submission_judge_result
FROM voj_submissions  s1
INNER JOIN (
    SELECT MAX(submission_id) AS max_submission_id, problem_id
    FROM voj_submissions
    WHERE uid = 1000
    AND problem_id >= 1000
    AND problem_id < 1010
    GROUP BY problem_id
) s2 
ON s1.problem_id = s2.problem_id
AND s1.submission_id = s2.max_submission_id

这在MariaDB和MySQL中均有效.

标签：sql-order-by,mariadb,mysql,group-by
来源： https://codeday.me/bug/20191120/2043503.html