sql - 跨移动窗口计算字符串值

Question

我正在尝试计算移动窗口中字符串值的出现次数。具体来说，我想计算前 3 行中每个字符串值的出现次数——不包括该行

我的数据看起来像这样:

id |  color
---+---------
 0 |  'blue'
 : |      :
 6 | 'green'
 7 |  'blue'
 8 | 'green'
 9 |   'red'
10 |   'red'

我正试图得到这样的东西:

id | n_red | n_blue | n_green 
---+-------+--------+---------
 : |     : |      : |       :
 9 |     0 |      1 |       2
10 |     1 |      1 |       1

其中数据是前 3 行中每个“红色”、“蓝色”和“绿色”的出现次数(例如，在 id 10 之前的 3 行中有 1 个红色， 1 个蓝色和 1 个绿色)。

我认为我应该能够使用窗口函数来完成此操作，但还没有完全解决。

select
    sum(red) over(order by id rows between 3 preceding and 1 preceding) end as n_red,
    sum(blue) over(order by id rows between 3 preceding and 1 preceding) end as n_blue,
    sum(green) over(order by id rows between 3 preceding and 1 preceding) end as n_green
from (select id,
        case when color = 'red' then 1 else 0 end as red,
        case when color = 'blue' then 1 else 0 end as blue,
        case when color = 'green' then 1 else 0 end as green
    from color_table) as dummy_color_table

这似乎可行，但不是很简洁。有没有人有编写这些类型的查询的经验，谁可以改进它？

Answer 1

您可以利用 Postgres 中的 :::

select sum((color = 'red')::int) over (order by id rows between 3 preceding and 1 preceding) end as n_red,
       sum((color = 'blue')::int) over (order by id rows between 3 preceding and 1 preceding) end as n_blue,
       sum((color = 'green')::int) over (order by id rows between 3 preceding and 1 preceding) end as n_green
from color_table;