我想找到每种客户类型每周的销售额百分比。我可以看到数学运算,但无法弄清楚如何编写查询。
SELECT
c.customerType as 'Customer Type',
DATEPART(WEEK, o.orderDate) as 'Week of the year',
COUNT(c.customerType) as 'Number of sales'
FROM
[dbo].[Order] o
JOIN
Customer c ON c.id = o.customerId
GROUP BY
c.customerType, DATEPART(WEEK, o.orderDate)
此查询输出按客户类型分组的每个销售的计数。
CustomerType Week Number of Sales
------------------------------------
Cash 36 248
Corporate 36 10
Personal 36 5
Cash 37 113
Corporate 37 3
Personal 37 2
Cash 38 136
Corporate 38 7
Personal 38 2
Cash 39 138
Corporate 39 4
Personal 39 3
答案 0 :(得分:0)
您可以包装查询并使用窗口函数:
select
t.*,
(100.0 * [Number of sales])
/(sum([Number of sales]) over(partition by [Week of the year])
[Percent of Total Sales]
from (
select
c.customerType as [Customer Type],
datepart(week, o.orderDate) as [Week of the year],
count(c.customerType) as [Number of sales],
from [dbo].[Order] o
join Customer c ON c.id = o.customerId
group by c.customerType, datepart(week, o.orderDate), datepart(year, o.orderDate)
) t
注意:
在SQLServer中,比起引号,最好使用方括号来定义标识符(引号通常保留给字符串)
.0
中的100.0
很重要:它强制SQLServer执行十进制除法(默认情况下,它将进行整数除法,这不是您想要的)
我将年份添加到组的定义中;如果您的数据分散了几年,您可能不希望将不同年份的同一周一起计算
侧面说明:SQLServer在混合窗口函数和聚合方面非常灵活。所以可能也可以:
select
c.customerType as [Customer Type],
datepart(week, o.orderDate) as [Week of the year],
count(c.customerType) as [Number of sales],
(100.0 * count(c.customerType))
/ (sum(count(c.customerType)) over(partition by datepart(week, o.orderDate)))
as [Percent of Total Sales]
from [dbo].[Order] o
join Customer c ON c.id = o.customerId
group by c.customerType, datepart(week, o.orderDate), datepart(year, o.orderDate)