获得 20 个最常见的数字对 (2/2) 和三元组 (3/3)

Question

我正在尝试获取最常见的数字对和三元组 (3/3)，我的表格如下所示：

+----+------+------+------+------+------+------+------+------+------+------+
| id | nr1  | nr2  | nr3  | nr4  | nr5  | nr6  | nr7  | nr8  | nr9  | nr10 |
+----+------+------+------+------+------+------+------+------+------+------+
| 1  | 1    | 39   | 19   | 23   | 28   | 80   | 3    | 42   | 60   | 32   |
+----+------+------+------+------+------+------+------+------+------+------+
| 2  | 43   | 18   | 3    | 24   | 29   | 33   | 15   | 1    | 61   | 80   |
+----+------+------+------+------+------+------+------+------+------+------+
| 3  | 11   | 25   | 33   | 2    | 30   | 3    | 1    | 44   | 62   | 78   |
+----+------+------+------+------+------+------+------+------+------+------+

我想知道我所有行中出现频率最高的前 3 个数字对或三组数字。

例子：

1,3（3次）

1,80（2 次）

3,80（2 次） 1,3,80（2次）

我可以尝试按 1、2、3 之类的顺序添加数字，然后从数据库中提取它们，但我想出的脚本仍然很糟糕，需要几个小时才能检查 10000 行

欢迎任何想法.. 非常感谢。

Answer 1

你需要对你的表进行 unpivot，但是 mysql 没有 unpivot 功能所以你可以这样做

SQL Fiddle Demo

CREATE TABLE unpivot
SELECT *
FROM ( 
        SELECT id, nr1 as n_value FROM tuple union all 
        SELECT id, nr2 as n_value FROM tuple union all 
        SELECT id, nr3 as n_value FROM tuple union all 
        SELECT id, nr4 as n_value FROM tuple union all 
        SELECT id, nr5 as n_value FROM tuple union all 
        SELECT id, nr6 as n_value FROM tuple union all 
        SELECT id, nr7 as n_value FROM tuple union all 
        SELECT id, nr8 as n_value FROM tuple union all 
        SELECT id, nr9 as n_value FROM tuple union all 
        SELECT id, nr10 as n_value FROM tuple
     ) as T

现在查找与自身连接的对数。

SELECT n1, n2, count(*) as total
FROM 
    (
    SELECT up1.n_value as n1, up2.n_value as n2
    FROM unpivot up1
    JOIN unpivot up2
      ON up1.`id` = up2.`id`        
     AND up1.n_value < up2.n_value  
   ) T
GROUP BY n1, n2
ORDER BY total desc
LIMIT 3;

对于三胞胎，您可以加入餐桌三次

SELECT n1, n2, n3, count(*) as total
FROM 
    (
    SELECT up1.n_value as n1, up2.n_value as n2, up3.n_value as n3
    FROM unpivot up1
    JOIN unpivot up2
      ON up1.`id` = up2.`id`        
     AND up1.n_value < up2.n_value  
     JOIN unpivot up3
      ON up2.`id` = up3.`id`        
     AND up2.n_value < up3.n_value  
   ) T
GROUP BY n1, n2, n3
ORDER BY total desc
LIMIT 3;

更新：

我在 postgresql 上做了测试

创建 50k 行，随机值从 1 到 90

创建索引后，查询只需 2 秒即可完成。