SQL 显示 SCD 类型 2 维度的时间趋势的查询 table

Question

Table 如下：

pk,EmployeeNo,Building,ValidFrom,ValidTo. 
1, 1, a, 2000-01-01, 2008-06-01
2, 1, b, 2008-06-01, 2010-06-01
3, 1, c, 2010-06-01, 2011-08-01
4, 2, a, 2000-01-01, 2008-06-01
5, 2, b, 2008-06-01, 2010-06-01
6, 2, c, 2010-06-01, 2011-08-01

我是一个相对新手 SQL 的开发人员，我使用的是 MS SQL 服务器。我想设计一个 SQL 查询，它可以采用上面的 table - 使用类型 2 构建缓慢变化的维度（具有有效日期和有效日期） - 并生成 table随着时间的推移，每日趋势客户计数。

输出如下：

Date, Employee Count
1/1/2000, 2
1/2/2000, 2

Answer 1

您可以使用聚合和累积总和 -- 在对数据进行逆透视后：

with d as (
      select validfrom as dte, 1 as inc
      from t
      union all
      select validto, -1
      from t
     )
select dte, sum(sum(inc)) over (order by dte)
from d
group by dte
order by dte;

注意：这假定不包括 validto 日期。如果你想把它包括在内，那就加“1”。

编辑：

如果您只需要这两个日期的结果，则使用：

select v.dte,
       (select count(*) 
        from t
        where t.validfrom <= v.dte and t.validto >= t.validto
       ) as cnt
from (values ('2020-01-01'), ('2020-01-02')) v(dte);

Answer 2

以下 SQL 似乎有效。我假设员工在 ValidTo 日期之前有效，但不是在 ValidTo 日期之前有效，以避免在 ValidFrom 和 ValidTo 重叠时重复计算这些日期；如果这个假设不正确，您需要修改 INNER JOIN 逻辑。

    --Get the min and max dates in the employee table
DECLARE @todate date, @fromdate date
SELECT @fromdate= min(ValidFrom) from [POC].[dbo].[EmployeeSCD2];
SELECT @todate= max(validto) from [POC].[dbo].[EmployeeSCD2];

-- Generate a list of all the dates between the two dates
WITH calendar (FromDate) AS (
    SELECT @fromdate AS FromDate
    UNION ALL
    SELECT DATEADD(day, 1, FromDate)
    FROM Calendar
    WHERE FromDate < @todate
)
-- Join the list of dates to the employee table - gives a record per date per employee active on that date
SELECT CAL.FromDate, COUNT(EMP.EMPLOYEENO) 'Employee Count' 
FROM CALENDAR CAL
INNER JOIN [POC].[dbo].[EmployeeSCD2] EMP ON CAL.FromDate >= EMP.ValidFrom AND CAL.FromDate < EMP.ValidTo
group by cal.FromDate
OPTION (MAXRECURSION 0) -- Without this parameter the recursion stops after 100 loops
;

Answer 3

一个简单的方法是建立一个日期 table，然后计算在该特定日期有多少员工（或客户）。

;with t as (   -- this is your table/Dimension
    select pk,EmployeeNo,Building,ValidFrom,ValidTo 
    from (values (1, 1, 'a', '2000-01-01', '2008-06-01'),
                 (2, 1, 'b', '2008-06-01', '2010-06-01'),
                 (3, 1, 'c', '2010-06-01', '2011-08-01'),
                 (4, 2, 'a', '2000-01-01', '2008-06-01'),
                 (5, 2, 'b', '2008-06-01', '2010-06-01'),
                 (6, 2, 'c', '2010-06-01', '2011-08-01')
        )t (pk,EmployeeNo,Building,ValidFrom,ValidTo)
)
, dates as ( -- this is a recursive query, building a date table from 2000-01-01 up to today
    select convert(date, '2000-01-01') as d
    union all
    select DATEADD(d,1,d) 
    from dates 
    where d < getdate()-1
)
select d.d, count(1) as [Employe count]  
from dates d
inner join t on d.d between t.ValidFrom and t.ValidTo -- will join only valid employee
group by d.d 
option (maxrecursion 0)  -- since our dates table recurse over more than 100 times and the dataset is fairly small, you need this option

SQL 显示 SCD 类型 2 维度的时间趋势的查询 table

SQL query to show time trend of a SCD Type 2 dimension table

sql

sql-server

database-design

etl