自季度开始以来的 SQL 查询计数

Question

我正在尝试确定自季度开始以来查询数据集以仅计算个人一次的最佳方式。这是数据集：

Year     |    Qtr    |    Week    |    Name
2021          1           1            John Smith
2021          1           3            John Smith
2021          2           14           John Smith
2021          1           1            Julie Johnson
2021          1           2            Julie Johnson
2021          1           3            John Jacobs

我需要能够返回一个数据集，该数据集告诉我每周的唯一身份人数以及季度至今的唯一身份人数：

结果集：

Year     |    Qtr    |    Week    |    Week Count    |    Rolling QTR Count
2021          1           1            2                  2   
2021          1           2            1                  2
2021          1           3            2                  3
2021          2           14           1                  1

我曾尝试使用窗口函数进行计数，但我无法弄清楚如何仅在滚动时间范围内计算某人。

Answer 1

只需在名称出现在每个季度的第一周添加一个标志，并将该标志用于季度列：

select year, qtr, week, count(*) as week_count,
       sum(sum(case when seqnum = 1 then 1 else 0 end)) over
           (partition by year, qtr order by week) as quarter_count
from (select d.*,
             row_number() over (partition by year, qtr, name order by week) as seqnum
      from dataset d
     ) d;

Answer 2

这里有一个解决您问题的方法（虽然我会尽量优化它）：

架构：

create table mytable(Year int, Qtr int,Week int, Name varchar(50));
insert into mytable values(2021,1,1,            'John Smith');
insert into mytable values(2021,1,3 ,           'John Smith');
insert into mytable values(2021,2,14 ,          'John Smith');
insert into mytable values(2021,1,1   ,         'Julie Johnson');
insert into mytable values(2021,1,2    ,        'Julie Johnson');
insert into mytable values(2021,1,3     ,       'John Jacobs');

查询：

with RollingQtyCount as (
select   m.year,m.qtr,m.week,count(name) over (partition by m.year,m.qtr,m.week) weekcount,
count(t.distinct_names)over(partition by m.year,m.qtr order by m.week rows between unbounded preceding and current row) [Rolling QTR Count]from mytable m

left join 
(
     select year,qtr,week, name distinct_names from mytable m where 
not exists (select name from mytable mt where m.name = mt.name and m.year = mt.year and m.qtr = mt.qtr and m.week > mt.week)

) t on m.year=t.year and m.qtr=t.qtr and m.week=t.week and m.name=t.distinct_names 
)
select year,qtr,week ,max(weekcount)weekcount,max([Rolling QTR Count]) [Rolling QTR Count] from RollingQtyCount
group by year,qtr,week
GO

year	qtr	week	weekcount	Rolling QTR Count
2021	1	1	2	2
2021	1	2	1	2
2021	1	3	2	3
2021	2	14	1	1

db小提琴here