Hive SQL 中按年、月、位置、状态的累计总和

Question

我想根据其他列在recorrencia列中使用 Hive SQL 进行累积计数。

 +------------+---------+-------+--------------+--+
| t.ano_mes  | t.site  | t.uf  | recorrencia  |
+------------+---------+-------+--------------+--+
| 202001     | 174     | AM    | 1            |
| 202002     | 174     | AM    | 1            |
| 202003     | 174     | AM    | 1            |
| 202004     | 174     | AM    | 1            |
| 202005     | 174     | AM    | 1            |
| 202006     | 174     | AM    | 1            |
| 202007     | 174     | AM    | 1            |
| 202008     | 174     | AM    | 1            |
| 202005     | 1JN     | SP    | 1            |
| 202006     | 1JN     | SP    | 1            |
| 202005     | 1LJ     | SP    | 1            |
| 202009     | 1LJ     | SP    | 1            |
| 202001     | 1RG     | SP    | 1            |
| 202002     | 1RG     | SP    | 1            |
| 202003     | 1RG     | SP    | 1            |
| 202004     | 1RG     | SP    | 1            |
| 202005     | 1RG     | SP    | 1            |
| 202006     | 1RG     | SP    | 1            |
| 202007     | 1RG     | SP    | 1            |

期望输出

+------------+---------+-------+--------------+--------+
| t.ano_mes  | t.site  | t.uf  | recorrencia  |cum_rec
+------------+---------+-------+--------------+--------+
| 202001     | 174     | AM    | 1            |1
| 202002     | 174     | AM    | 1            |2
| 202003     | 174     | AM    | 1            |3
| 202004     | 174     | AM    | 1            |4
| 202005     | 174     | AM    | 1            |5
| 202006     | 174     | AM    | 1            |6
| 202007     | 174     | AM    | 1            |7
| 202008     | 174     | AM    | 1            |8
| 202005     | 1JN     | SP    | 1            |1
| 202006     | 1JN     | SP    | 1            |2
| 202005     | 1LJ     | SP    | 1            |1
| 202009     | 1LJ     | SP    | 1            |2
| 202001     | 1RG     | SP    | 1            |1
| 202002     | 1RG     | SP    | 1            |2
| 202003     | 1RG     | SP    | 1            |3
| 202004     | 1RG     | SP    | 1            |4
| 202005     | 1RG     | SP    | 1            |5
| 202006     | 1RG     | SP    | 1            |6
| 202007     | 1RG     | SP    | 1            |7

我已经尝试了很多函数，如COUNT(*) OVER (t.ano_mes)和COUNT(*) OVER (t.site)但它运行总和直到表结束，并且不作为t.site重新启动变化。

一旦t.site更改，计数器应重新启动。

Answer 1

那将是：

sum(recorrencia) over(partition by t.site order by t.ano_mes) as cum_rec

partition by子句会在每次站点更改时重置总和。

请注意，如果recorrencia始终为1 ，如您的示例数据所示，则row_number()就足够了：

row_number() over(partition by t.site order by t.ano_mes) as cum_rec