为A列的每个值选择N列B的最频繁值

Question

用如下的MySQL表：

id | colA | colB
...| 1    | 13
...| 1    | 13
...| 1    | 12
...| 1    | 12
...| 1    | 11
...| 2    | 78
...| 2    | 78
...| 2    | 78
...| 2    | 13
...| 2    | 13
...| 2    | 9

对于每个值colA我想找到N个最频繁的值colB 。

N = 2的示例结果：

colA | colB
1    | 13
1    | 12
2    | 78
2    | 13

我能够使用以下方法获得colA和colB所有唯一组合及其频率：

SELECT colA, colB, COUNT(*) AS freq FROM t GROUP BY colA, colB ORDER BY freq DESC;

结果示例：

colA | colB | freq
1    | 13   | 2
1    | 12   | 2
1    | 11   | 1
2    | 78   | 3
2    | 13   | 2
2    | 9    | 1

但是我很难为colA每个值而不是整个表应用LIMIT 。

基本上就像如何在每个ID组的列中选择最频繁的值？ ，仅适用于MySQL而非PostgreSQL。

我目前正在使用MariaDB 10.1。

Answer 1

如果可以，请使用窗口功能：

SELECT colA, colB, freq
FROM (SELECT colA, colB, COUNT(*) AS freq,
             DENSE_RANK() OVER (PARTITION BY colA ORDER BY COUNT(*) DESC) as seqnum
      FROM t
      GROUP BY colA, colB 
     ) ab
WHERE seqnum <= 2;

请注意，根据您对待领带的方式，可能需要DENSE_RANK() ， RANK()或ROW_NUMBER() 。 如果有5个colB值具有最高的两个等级，则DENSE_RANK()将返回所有五个。

如果只需要两个值，则使用ROW_NUMBER() 。

Answer 2

您可能可以为此使用几个CTE，例如：

WITH counts AS (
   SELECT colA, colB, COUNT(*) AS freq FROM t GROUP BY colA, colB ORDER BY freq DESC
), most_freq AS (
   SELECT colA, max(freq) FROM counts GROUP BY colA
)
   SELECT counts.*
     FROM counts
     JOIN most_freq ON (counts.colA = most_freq.colA 
                        AND counts.freq = most_freq.freq);