下面使用用户定义变量的方法比使用相关子查询的方法性能更好。对你来说 temp 变量被用来解释计算逻辑，它也被输出。你可以忽略这一点。
您可以尝试以下查询（如果需要，可以添加更多解释）：
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，我已订购 seller_id 先是然后 transaction_date )

| 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       |

db fiddle视图

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;

做这样的事吗？

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;

样品

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]>