sqlfiddle.com 5.5.30 和 MariaDB 5.5.31 中的不同结果

sqlfiddle:http://sqlfiddle.com/#!2/9a8b3/1

从小提琴中获取结构和数据并进行查询， 导入到我的 MariaDB 5.5.31 中，我得到不同的结果：

sqlfiddle

PID  NAME       LEAGUEPOINTS        TOTALLEAGUEPOINTS
2   Peter   16,13,9,4,2            44
1   Daniel  3425,543,234,43,29,22,21,21,19,17,13,12,12,12,11,9,9,9,8,7      4476

mariadb

pid  name    leaguepoints       totalleaguepoints   
2   Peter   16,13,9,4,2             44
1   Daniel  3425,543,234,43,29,22,21,21,19,17,13,12,12,12,11,9,9,9,8,7,7,6,5,5,4,4,4,3,3,2,1    4520

Query:

SELECT                
    p.pid,
    p.name,   
    GROUP_CONCAT( gC.leaguepoints ORDER BY leaguepoints DESC ) AS leaguepoints, 
    SUM(gC.leaguepoints) AS totalleaguepoints
FROM test_golf_player p
LEFT JOIN 
(
    SELECT pid, leaguepoints, @Sequence:=IF(@PrevPid = pid, @Sequence + 1, 0) AS aSequence, @PrevPid := pid
    FROM
    (
        SELECT pid, leaguepoints
        FROM test_golf_card 
        ORDER BY pid, leaguepoints DESC
    ) Sub1
    CROSS JOIN (SELECT @PrevPid := 0, @Sequence := 0) Sub2
) gC
ON p.pid = gC.pid AND aSequence < 20
GROUP BY p.pid
ORDER BY p.name DESC 

知道为什么吗？

恐怕我手头没有 MariaDB，但是你可以尝试以下操作来看看用户变量是如何输出的：-

SELECT  *
FROM test_golf_player p
LEFT JOIN 
(
    SELECT pid, leaguepoints, @Sequence:=IF(@PrevPid = pid, @Sequence + 1, 0) AS aSequence, @PrevPid := pid
    FROM
    (
        SELECT pid, leaguepoints
        FROM test_golf_card 
        ORDER BY pid, leaguepoints DESC
    ) Sub1
    CROSS JOIN (SELECT @PrevPid := 0, @Sequence := 0) Sub2
) gC
ON p.pid = gC.pid 
ORDER BY p.name DESC 

编辑-做了一些调查查看结果，MariaDB 似乎忽略了子查询中的 ORDER BY。因此，序列号是随机顺序的，并且当 pid 改变时也会重置（由于顺序不固定，它会随机重置）。谷歌了一下，看来这是 MariaDB 故意的一个功能。 SQL 标准将表定义为无序的行集，并且子选择被视为表，因此忽略 order by -https://kb.askmonty.org/en/why-is-order-by-in-a-from-subquery-ignored/ .

这是一个有点缺点。不确定是否有解决办法，因为我目前想不出解决办法。对于要处理的原始问题，我认为有必要使用相关的子选择，这可能效率不高。