MySQL 将多个子查询优化为嵌套内部连接？

Question

我有 3 个表；契约(Contract)、经销商和用户。

用户有很多经销商，经销商有很多契约(Contract)，但契约(Contract)与用户没有直接关联。

我正在尝试构建一份报告，以获取过去 12 个月按用户分组的每月已完成契约(Contract)数。

到目前为止，我已经构建了一个非常慢的多子查询:SQL Fiddle

SELECT *,
  ( SELECT count(*) FROM contracts
     WHERE
       dealer_id IN
         ( SELECT id FROM dealers WHERE user_id = User.id )
       AND status = 'Paid'
       AND completion_date BETWEEN
         '2012-08-01 00:00:00' AND '2012-08-31 23:59:59'
  ) AS Aug_2012,

  ( SELECT count(*) FROM contracts
     WHERE
       dealer_id IN
         ( SELECT id FROM dealers WHERE user_id = User.id )
       AND status = 'Paid'
       AND completion_date BETWEEN
         '2012-09-01 00:00:00' AND '2012-09-30 23:59:59'
  ) AS Sep_2012
FROM users AS User
  WHERE
    id IN( SELECT user_id FROM dealers WHERE active = 1 AND user_id IS NOT NULL GROUP BY user_id )
    AND id != 1
  ORDER BY User.name ASC

我想使用这样的东西代替每个月选择的子查询:

COUNT(*) as last_12_months,
SUM(case when MONTH(completion_date) = 8 then 1 else 0 end) as Aug_2012,
SUM(case when MONTH(completion_date) = 9 then 1 else 0 end) as Sep_2012,
etc.

由于我要返回多个列，因此我必须对其进行重组，但我不确定如何进行重组。如果我使用 INNER JOIN，我在什么子句上加入？

下面是基于 Mikhail 的回答的最终查询:

SELECT
User.*,
SUM(case when MONTH(completion_date) = 8 then 1 else 0 end) AS Aug_2012,
SUM(case when MONTH(completion_date) = 9 then 1 else 0 end) AS Sep_2012,
SUM(case when MONTH(completion_date) = 10 then 1 else 0 end) AS Oct_2012,
SUM(case when MONTH(completion_date) = 11 then 1 else 0 end) AS Nov_2012,
SUM(case when MONTH(completion_date) = 12 then 1 else 0 end) AS Dec_2012,
SUM(case when MONTH(completion_date) = 1 then 1 else 0 end) AS Jan_2013,
SUM(case when MONTH(completion_date) = 2 then 1 else 0 end) AS Feb_2013,
SUM(case when MONTH(completion_date) = 3 then 1 else 0 end) AS Mar_2013,
SUM(case when MONTH(completion_date) = 4 then 1 else 0 end) AS Apr_2013,
SUM(case when MONTH(completion_date) = 5 then 1 else 0 end) AS May_2013,
SUM(case when MONTH(completion_date) = 6 then 1 else 0 end) AS Jun_2013,
SUM(case when MONTH(completion_date) = 7 then 1 else 0 end) AS Jul_2013,
SUM(case when completion_date BETWEEN '2012-08-01 00:00:00' AND '2013-07-31 23:59:59' then 1 else 0 end) as last_12_months
FROM users AS User
LEFT OUTER JOIN 
( 
    SELECT id, user_id FROM dealers 
    WHERE active = 1 AND user_id IS NOT NULL
) AS Dealer ON User.id = Dealer.user_id
LEFT OUTER JOIN
(
    SELECT completion_date, status, dealer_id FROM contracts
    WHERE completion_date BETWEEN '2012-08-01 00:00:00' AND '2013-07-31 23:59:59' AND status = 'Paid' AND cancelled = 0
) AS Contract on Dealer.id = Contract.dealer_id
WHERE
    User.id IN 
    (
        SELECT user_id FROM dealers
        WHERE active = 1 AND user_id IS NOT NULL
        GROUP BY user_id
    )
GROUP BY
    User.id order by User.name asc

这大约快了 4 倍。

Answer 1

试试这个:

select
    User.id, User.name,
    sum(case when MONTH(completion_date) = 8 and Year(completion_date)=2012 then 1 else 0 end) as Aug_2012,
    sum(case when MONTH(completion_date) = 9 and Year(completion_date)=2012 then 1 else 0 end) as Sep_2012,
    sum(case when MONTH(completion_date) = 10 and Year(completion_date)=2012 then 1 else 0 end) as Oct_2012,
    sum(case when MONTH(completion_date) = 11 and Year(completion_date)=2012 then 1 else 0 end) as Nov_2012,
    sum(case when MONTH(completion_date) = 12 and Year(completion_date)=2012 then 1 else 0 end) as Dec_2012,
    sum(case when MONTH(completion_date) = 1 and Year(completion_date)=2013 then 1 else 0 end) as Jan_2012,
    sum(case when MONTH(completion_date) = 2 and Year(completion_date)=2013 then 1 else 0 end) as Feb_2012,
    sum(case when MONTH(completion_date) = 3 and Year(completion_date)=2013 then 1 else 0 end) as Mar_2012,
    sum(case when MONTH(completion_date) = 4 and Year(completion_date)=2013 then 1 else 0 end) as Apr_2012,
    sum(case when MONTH(completion_date) = 5 and Year(completion_date)=2013 then 1 else 0 end) as May_2012,
    sum(case when MONTH(completion_date) = 6 and Year(completion_date)=2013 then 1 else 0 end) as Jun_2012,
    sum(case when MONTH(completion_date) = 7 and Year(completion_date)=2013 then 1 else 0 end) as Jul_2012
from users AS User
left outer join dealers on
    User.id=dealers.user_id
left outer join contracts on
    dealers.id=contracts.dealer_id
group by
    User.id,
    contracts.status
having
    contracts.status='Paid'
order by
    User.name asc;