为什么 SQL 不允许多个子查询使用 WITH 子句？

Question

为什么 SQL 只允许嵌套子查询？

例如，拿这个问题

• 找出每个职业中评分最高的用户。

表名是ratings ，有列

1. 用户身份
2. 职业
3. 评分

在 Postgres 或 Bigquery 中，我会做

with ratings_by_user as (
select occupation, user_id, count(*) num_ratings
from ratings
group by 1,2
),

max_ratings_by_occupation as (
select occupation, max(num_ratings) as max_ratings
from ratings_by_user
group by 1
),

select occupation, user_id
from ratings_by_user
inner join max_ratings_by_occupation
using (occupation)
where num_ratings = max_ratings

但我不确定如何在 SQL 中执行此操作，我需要将所有子查询嵌套在一个块中。 这是我在 SQL 中的尝试，但它不起作用。

select occupation, user_id, count(*) as num_ratings
from ( 
    select occupation, max(num_ratings) max_ratings 
    from ( 
        select occupation, user_id, count(*) num_ratings
        from users
        group by 1,2
        ) as ratings_table
    group by 1
    ) as max_ratings_table
)
inner join ratings on ratings.occupation = max_ratings_table.occupation
where max_ratings = num_ratings

任何人都可以启发我如何在 SQL 中使用相同样式的 Postgres / Bigquery 我希望按顺序处理我的子查询？ 我只是发现很难在一大块中解决复杂的问题。

非常感谢您的参与。

Answer 1

根据https://dev.mysql.com/doc/refman/8.0/en/with.html和MySQL“WITH”子句- WITH仅在 MySQL 8+ 上受支持。 确保您使用的是适当版本的 MySQL

转换为嵌套版本并不难。 我们使用您的工作 sql：

with ratings_by_user as (
select occupation, user_id, count(*) num_ratings
from ratings
group by 1,2
),

max_ratings_by_occupation as (
select occupation, max(num_ratings) as max_ratings
from ratings_by_user
group by 1
),

select occupation, user_id
from ratings_by_user
inner join max_ratings_by_occupation
using (occupation)
where num_ratings = max_ratings

我们复制所有内容，包括 WITH 的括号，并在使用别名之前将其粘贴。

步骤 1，剪切 ratings_by_user 并将其粘贴到使用 ratings_by_user 的任何地方（两次）

--cut from here
with ratings_by_user as ,

max_ratings_by_occupation as (
select occupation, max(num_ratings) as max_ratings
from 
  --paste to here
  (
    select occupation, user_id, count(*) num_ratings
    from ratings
    group by 1,2
  ) ratings_by_user
group by 1
),

select occupation, user_id
from
--and also paste to here 
(
  select occupation, user_id, count(*) num_ratings
  from ratings
  group by 1,2
) ratings_by_user
inner join max_ratings_by_occupation
using (occupation)
where num_ratings = max_ratings

第 2 步，剪切 max_ratings_by_occupation 并将其粘贴到使用它的位置：

with ratings_by_user as ,

--cut from here
max_ratings_by_occupation as ,

select occupation, user_id
from
(
  select occupation, user_id, count(*) num_ratings
  from ratings
  group by 1,2
) ratings_by_user
inner join 

--paste to here
(
  select occupation, max(num_ratings) as max_ratings
  from 
  (
    select occupation, user_id, count(*) num_ratings
    from ratings
    group by 1,2
  ) ratings_by_user
  group by 1
) max_ratings_by_occupation

using (occupation)
where num_ratings = max_ratings

第3步，清理空的withs

select occupation, user_id
from
(
  select occupation, user_id, count(*) num_ratings
  from ratings
  group by 1,2
) ratings_by_user
inner join 
(
  select occupation, max(num_ratings) as max_ratings
  from 
  (
    select occupation, user_id, count(*) num_ratings
    from ratings
    group by 1,2
  ) ratings_by_user
  group by 1
) max_ratings_by_occupation

using (occupation)
where num_ratings = max_ratings

这将是优化/重写的开始。 棘手的部分是它使用 ratings_by_user 两次，因此在步骤 1 中需要两次粘贴

您的重新格式化尝试没有成功，因为您试图在外部级别使用仅存在于内部级别的结果集：

select occupation, user_id, count(*) as num_ratings
from 

( --max_ratings_table available inside these brackets
    select occupation, max(num_ratings) max_ratings 
    from ( 
        select occupation, user_id, count(*) num_ratings
        from users
        group by 1,2
        ) as ratings_table
    group by 1
    ) as max_ratings_table
  --end of max_ratings_table availability
) 

inner join ratings on ratings.occupation = max_ratings_table.occupation
--                                         ^^^^^^^^^^^^^^^^^
--                                       mrt not available here
where max_ratings = num_ratings