SQL

Question

我无法弄清楚如何对不同的行求和，但使用每个分组的第一行作为起点。

我有一个 table 看起来像这样 :

+------+------+----------+---------------+----------+-------+
| Dim1 | Dim2 |   Date   | beg_Inventory | purchase | sales |
+------+------+----------+---------------+----------+-------+
| x    | y    | 1/1/2019 |           100 |       50 |    20 |
| x    | y    | 1/2/2019 |               |       70 |    80 |
| x    | y    | 1/3/2019 |               |       40 |    60 |
| x    | y    | 1/4/2019 |               |       30 |    50 |
| x    | y    | 1/5/2019 |               |      100 |    10 |
| x    | z    | 1/1/2019 |            65 |       10 |    50 |
| x    | z    | 1/2/2019 |               |       20 |     5 |
| x    | z    | 1/3/2019 |               |       40 |     5 |
+------+------+----------+---------------+----------+-------+

我想要这样的结果：

+------+------+----------+---------------+----------+-------+------------+
| Dim1 | Dim2 |   Date   | beg_Inventory | purchase | sales | ending_inv |
+------+------+----------+---------------+----------+-------+------------+
| x    | y    | 1/1/2019 |           100 |       50 |    20 |        130 |
| x    | y    | 1/2/2019 |               |       70 |    80 |        120 |
| x    | y    | 1/3/2019 |               |       40 |    60 |        100 |
| x    | y    | 1/4/2019 |               |       30 |    50 |         80 |
| x    | y    | 1/5/2019 |               |      100 |    10 |        170 |
| x    | z    | 1/1/2019 |            65 |       10 |    50 |         25 |
| x    | z    | 1/2/2019 |               |       20 |     5 |         40 |
| x    | z    | 1/3/2019 |               |       40 |     5 |         75 |
+------+------+----------+---------------+----------+-------+------------+

期末库存为 beg_inventory + 采购 - 销售，对于每个 dim1 和 dim2 分组的第一行，即第一行为 100+50-20=130 但是，在第2行中，它必须使用我们计算的130并使用下一行的购买和销售得到130 + 70-80 = 120等等，按dim1和dim2分组并按日期排序。

谢谢。

Answer 1

尝试使用 first_value() 获取 window 中 beg_inventory 的第一个值，然后添加 windowed sum() 的采购和销售

SELECT dim1,
       dim2,
       date,
       beg_inventory,
       purchase,
       sales,
       first_value(beg_inventory) OVER (PARTITION BY dim1,
                                                     dim2
                                        ORDER BY date)
       +
       sum(purchase - sales) OVER (PARTITION BY dim1,
                                                dim2
                                   ORDER BY date) ending_inv
       FROM elbat
       ORDER BY dim1,
                dim2,
                date;

Answer 2

所有行的公式对我来说都一样：

SELECT t.*,
       SUM(COALESCE(beg_Inventory, 0) + purchase - sales) OVER (PARTITION BY dim1, dim2 ORDER BY date) as ending_inv
FROM t
ORDER BY dim1, dim2, date;

SQL - 如何对不同行求和，同时第一行有不同的方程式

SQL - how to sum across different rows while having a different equation for the first row

teradata