SQL Group-By 最常见

Question

假设我有下表

+-----------------------+----------------+
| PERSON                | ATE            |
+-----------------------+----------------+
| Joe                   | Chicken        |
| Joe                   | Chicken        |
| Joe                   | Chicken        |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Joe                   | Chicken        |
| Joe                   | Tuna           |
| Joe                   | Chicken        |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Joe                   | Chicken        |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Joe                   | Chicken        |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Joe                   | Tuna           |
| Rob                   | Chicken        |
| Rob                   | Tuna           |
| Rob                   | Chicken        |
| Rob                   | Chicken        |
| Rob                   | Chicken        |
| Rob                   | Chicken        |
| Jonathan              | Tuna           |
| Jonathan              | Tuna           |
| Jonathan              | Chicken        |
| Jonathan              | Tuna           |
| Jonathan              | Chicken        |
+-----------------------+----------------+

我想为每个人找到最常吃的食物。所以，想要的结果是：

+-----------------------+----------------+
| PERSON                | ATE            |
+-----------------------+----------------+
| Joe                   | Tuna           |
| Rob                   | Chicken        |
| Jonathan              | Tuna           |
+-----------------------+----------------+

当然，我会 GROUP BY（人）。之后，我必须找到每组中最常吃的项目。如果它们是数字，那会很容易，因为我可以只使用 MAX 函数，但我不知道如何从一堆字符串中挑选出最受欢迎的字符串。

你将如何完成我在 SQL 中尝试做的事情？

Answer 1

您可以使用相关子查询：

select distinct t1.PERSON, (
    select ATE
    from myTable t2
    where t2.PERSON = t1.PERSON
    group by ATE
    order by count(*) desc
    limit 1
) as ATE
from myTable t1

如果您有关系，此查询将“随机”选择吃得最多的项目之一。

使用 MySQL 8 或 MariaDB 10.2（都还不稳定），您将能够使用 CTE（公共表表达式）

with t1 as (
    select PERSON, ATE, count(*) as cnt
    from myTable
    group by PERSON, ATE
), t2 as (
    select PERSON, max(cnt) as cnt
    from t1
    group by PERSON
)
    select *
    from t1
    natural join t2

在平局上，此查询可能会返回每个组 (PERSON) 的多行。