SQL - 如何获得大于不同列中当前值的 3 个前导值的总和答案

【问题标题】：SQL - How to get the sum of 3 leading values greater than the current value in a different columnSQL - 如何获得大于不同列中当前值的 3 个前导值的总和
【发布时间】：2021-08-26 06:00:36
【问题描述】：

我有一个数据表，如下所示

我想要的输出在最后一列中，它本质上是大于当前值的月份差异的 sum(value for next 3 leading value)，即第一行中的值 33 是 sum(Value) for month_diff>0 i.e. 1-3，类似地，第 7 行 22 是 @ 987654327@。以此类推。

它必须按 id 列进行分区

【问题讨论】：

标签： sql google-bigquery

【解决方案1】：

根据您的描述，您需要一个窗框规格：

select t.*,
       sum(value) over (partition by id1, id2
                        order by month_diff
                        rows between 1 following and 3 following
                       ) as desired_output
from t;

如果你想确定你实际上有三个值，那么你需要一些case 逻辑：

select t.*,
       (case when lead(id1, 3) over (partition by id1, id2 order by month_diff) is not null
             then sum(value) over (partition by id1, id2
                                   order by month_diff
                                   rows between 1 following and 3 following
                                  )
        end) as desired_output
from t;

【讨论】：

【解决方案2】：

考虑以下方法

select *, 
  if(count(*) over win = 3, sum(value) over win, 0) output
from `project.dataset.tabel`
window win as (
  partition by id1, id2 
  order by month_diff 
  rows between 1 following and 3 following
)

【讨论】：