我有一个名为transactions的表，其中包含卖方及其交易：销售，废物以及何时收到产品。结构基本上如下：
seller_id transaction_date quantity reason product unit_price
--------- ---------------- -------- ------ ------- ----------
1 2018-01-01 10 import 1 100.0
1 2018-01-01 -5 sale 1 100.0
1 2018-01-01 -1 waste 1 100.0
2 2018-01-01 -3 sale 4 95.5
我需要每个卖方的每日摘要，包括其销售，浪费和开始存货的价值。问题在于，初始库存是直到给定日期为止的累计数量总和（也包括给定日期的进口量）。我有以下查询：
SELECT
t.seller_id,
t.transaction_date,
t.SUM(quantity * unit_price) as amount,
t.reason as reason,
(
SELECT SUM(unit_price * quantity) FROM transactions
WHERE seller_id = t.seller_id
AND (transaction_date <= t.transaction_date)
AND (
transaction_date < t.transaction_date
OR reason = 'import'
)
) as opening_balance
FROM transactions t
GROUP BY
t.transaction_date,
t.seller_id
t.reason
查询有效，我得到了所需的结果。但是，即使在为外部查询和子查询创建索引之后，它也花费了太多时间（约30秒），因为opening_balance查询是一个依赖子查询，需要对每一行进行一次又一次的计算。 / p>
如何优化或重写此查询？
编辑：子查询有一个缺少WHERE条件的小错误，我对其进行了修复，但问题的实质是相同的。我用示例数据创建了一个小提琴来玩：
答案 0 :(得分：2)
使用用户定义的变量的以下方法可能比使用“相关子查询”更有效。在您的情况下，使用了一个temp变量来说明计算逻辑，该变量也将被输出。您可以忽略它。
您可以尝试以下查询（如果需要，可以添加更多说明）：
查询
SELECT dt.reason,
dt.amount,
@bal := CASE
WHEN dt.reason = 'import'
AND @sid <> dt.seller_id THEN dt.amount
WHEN dt.reason = 'import' THEN @bal + @temp + dt.amount
WHEN @sid = 0
OR @sid = dt.seller_id THEN @bal
ELSE 0
end AS opening_balance,
@temp := CASE
WHEN dt.reason <> 'import'
AND @sid = dt.seller_id
AND @td = dt.transaction_date THEN @temp + dt.amount
ELSE 0
end AS temp,
@sid := dt.seller_id AS seller_id,
@td := dt.transaction_date AS transaction_date
FROM (SELECT seller_id,
transaction_date,
reason,
Sum(quantity * unit_price) AS amount
FROM transactions
WHERE seller_id IS NOT NULL
GROUP BY seller_id,
transaction_date,
reason
ORDER BY seller_id,
transaction_date,
Field(reason, 'import', 'sale', 'waste')) AS dt
CROSS JOIN (SELECT @sid := 0,
@td := '',
@bal := 0,
@temp := 0) AS user_vars;
结果（请注意，我先按seller_id的顺序排序，然后按transaction_date的顺序排序）
| reason | amount | opening_balance | temp | seller_id | transaction_date |
| ------ | ------ | --------------- | ----- | --------- | ---------------- |
| import | 1250 | 1250 | 0 | 1 | 2018-12-01 |
| sale | -850 | 1250 | -850 | 1 | 2018-12-01 |
| waste | -100 | 1250 | -950 | 1 | 2018-12-01 |
| import | 950 | 1250 | 0 | 1 | 2018-12-02 |
| sale | -650 | 1250 | -650 | 1 | 2018-12-02 |
| waste | -450 | 1250 | -1100 | 1 | 2018-12-02 |
| import | 2000 | 2000 | 0 | 2 | 2018-12-01 |
| sale | -1200 | 2000 | -1200 | 2 | 2018-12-01 |
| waste | -250 | 2000 | -1450 | 2 | 2018-12-01 |
| import | 750 | 1300 | 0 | 2 | 2018-12-02 |
| sale | -600 | 1300 | -600 | 2 | 2018-12-02 |
| waste | -450 | 1300 | -1050 | 2 | 2018-12-02 |
答案 1 :(得分：1)
这样的事情吗？
SELECT s.* ,@balance:=@balance+(s.quantity*s.unit_price) AS opening_balance FROM (
SELECT t.* FROM transactions t
ORDER BY t.seller_id,t.transaction_date,t.reason
) s
CROSS JOIN ( SELECT @balance:=0) AS INIT
GROUP BY s.transaction_date, s.seller_id, s.reason;
示例
MariaDB [test]> select * from transactions;
+----+-----------+------------------+----------+------------+--------+
| id | seller_id | transaction_date | quantity | unit_price | reason |
+----+-----------+------------------+----------+------------+--------+
| 1 | 1 | 2018-01-01 | 10 | 100 | import |
| 2 | 1 | 2018-01-01 | -5 | 100 | sale |
| 3 | 1 | 2018-01-01 | -1 | 100 | waste |
| 4 | 2 | 2018-01-01 | -3 | 99.5 | sale |
+----+-----------+------------------+----------+------------+--------+
4 rows in set (0.000 sec)
MariaDB [test]> SELECT s.* ,@balance:=@balance+(s.quantity*s.unit_price) AS opening_balance FROM (
-> SELECT t.* FROM transactions t
-> ORDER BY t.seller_id,t.transaction_date,t.reason
-> ) s
-> CROSS JOIN ( SELECT @balance:=0) AS INIT
-> GROUP BY s.transaction_date, s.seller_id, s.reason;
+----+-----------+------------------+----------+------------+--------+-----------------+
| id | seller_id | transaction_date | quantity | unit_price | reason | opening_balance |
+----+-----------+------------------+----------+------------+--------+-----------------+
| 1 | 1 | 2018-01-01 | 10 | 100 | import | 1000 |
| 2 | 1 | 2018-01-01 | -5 | 100 | sale | 500 |
| 3 | 1 | 2018-01-01 | -1 | 100 | waste | 400 |
| 4 | 2 | 2018-01-01 | -3 | 99.5 | sale | 101.5 |
+----+-----------+------------------+----------+------------+--------+-----------------+
4 rows in set (0.001 sec)
MariaDB [test]>
答案 2 :(得分：0)
SELECT
t.seller_id,
t.transaction_date,
SUM(quantity) as amount,
t.reason as reason,
quantityImport
FROM transaction t
inner join
(
select sum(ifnull(quantityImport,0)) quantityImport,p.transaction_date,p.seller_id from
( /* subquery get all the date and seller distinct row */
select transaction_date ,seller_id ,reason
from transaction
group by seller_id, transaction_date
)
as p
left join
( /* subquery get all the date and seller and the import quantity */
select sum(quantity) quantityImport,transaction_date ,seller_id
from transaction
where reason='Import'
group by seller_id, transaction_date
) as n
on
p.seller_id=n.seller_id
and
p.transaction_date>=n.transaction_date
group by
p.seller_id,p.transaction_date
) as q
where
t.seller_id=q.seller_id
and
t.transaction_date=q.transaction_date
GROUP BY
t.transaction_date,
t.seller_id,
t.reason;