避免“不支持这种类型的相关子查询模式”的方法

Posted 2023-03-31

【中文标题】避免“不支持这种类型的相关子查询模式”的方法

【英文标题】：Ways to avoid "This type of correlated subquery pattern is not supported"

【发布时间】：2017-08-08 17:58:54

【问题描述】：

这就是我想要做的事情

select something1,something2,account_id,
(select u.organization_id
          from public.sfdc_contact sfdcc
          join public.users u on u.email=sfdcc.email
          where sfdcc.account_id=account_id
          group by u.organization_id
          order by count(*)
      limit 1
          )
from something

Redshift 提示我该错误，因为我试图按计数排序。但是我不能在子查询中有 2 列，有什么提示吗？

【问题讨论】：

改用连接（编写单独的子查询并将其连接到主查询）

【参考方案1】：

不完全确定此结构是否完全满足您的需求，但使用“window functions”如 ROW_NUMBER() OVER() 可用于从连接的派生表（子查询）提供单行。例如：

SELECT
      s.something1
    , s.something2
    , s.account_id
    , d2.organization_id
    , d2.cn
FROM something s
LEFT JOIN (
      SELECT
            organization_id
          , account_id
          , cn
          , ROW_NUMBER() OVER (PARTITION BY organization_id ORDER BY cn) rn
      FROM (
            SELECT
                  u.organization_id
                , sfdcc.account_id
                , COUNT(*) OVER (PARTITION BY u.organization_id, sfdcc.account_id) cn
            FROM public.sfdc_contact sfdcc
            JOIN public.users u ON u.email = sfdcc.email
            ) d1
      ) d2 ON s.account_id = d2.account_id and d2.rn = 1

使用 COUNT() OVER() 可能是不必要的，这可能更实用：

SELECT
      s.something1
    , s.something2
    , s.account_id
    , d2.organization_id
    , d2.cn
FROM something s
LEFT JOIN (
      SELECT
            organization_id
          , account_id
          , cn
          , ROW_NUMBER() OVER (PARTITION BY organization_id ORDER BY cn) rn
      FROM (
            SELECT
                  u.organization_id
                , sfdcc.account_id
                , COUNT(*) cn
            FROM public.sfdc_contact sfdcc
            JOIN public.users u ON u.email = sfdcc.email
            GROUP BY
                  u.organization_id
                , sfdcc.account_id
            ) d1
      ) d2 ON s.account_id = d2.account_id and d2.rn = 1

另请注意，如果您想要最高计数，请将 row_number 中使用的顺序更改为 DESCending：

, ROW_NUMBER() OVER (PARTITION BY organization_id ORDER BY cn DESC) rn