SQL，不完整的记录多长时间被按日分解

Question

我已经问过与此SQL Daily Outstanding Sales, Rolling Aggregate?类似的问题，但是我现在需要每天了解销售的持续时间。这将分为24小时（或一天）间隔，最多72小时。下面是数据集输出的一个示例，超时将在一个组内，但是日期必须是any日期，即使它不在数据集中也是如此（可以在2020-01-02上看到）第二行

Date         TimeOutstanding    VolumeOutstanding  
2020-01-01   Under24Hour        1                                   
2020-01-02   Under48Hour        1
2020-01-03   Under24Hour        3
2020-01-03   Under72Hour        1
2020-01-04   Under48Hour        3
2020-01-05   Under72Hour        2
2020-01-05   Over72Hour         1 

数据集

SaleID   Date         Outcome
1        2020-01-01   New
1        2020-01-01   Complete
2        2020-01-01   New
3        2020-01-03   New
4        2020-01-03   New
5        2020-01-03   New
2        2020-01-04   Complete
5        2020-01-04   Complete
3        2020-01-06   Complete
4        2020-01-07   Complete

Answer 1

首先，您可以用新的和完成的日期汇总每个Salesid。然后，生成您关心的日期。并使用条件聚合：

select d.date,
       count(*) as num_open,
       sum(case when new_date >= dateadd(day, -1, d) then 1 else 0
           end) as within_1day,
       sum(case when new_date < dateadd(day, -1, d)  and new_date >= dateadd(day, -3, d then 1 else 0
           end) as within_1_3days_ago,
       . . .
from (select distinct date from t) d left join
     (select salesid, min(date) as new_date,
             nullif(max(date), min(date)) as completed_date
      from t
      group by salesid
     ) t
     on new_date <= date and (date < completed_date or completed_date is null)
group by d.date