-3

我有一群经常变化的客户。每次群组更改时，都会将新客户的完整转储附加到表中并赋予新的 List_Name。Alpha、Beta、Gamma等。每个列表一般都会增加一些客户，删除一些客户，保留一些客户。我正在尝试创建一个简单的瀑布图来显示列表之间的变化。通常我只会 FULL OUTER JOIN Alpha<->Beta，然后 FULL OUTER JOIN Beta<->Gamma 等，以获取留下、离开和添加的客户。但是 List_Versions 增长了很多，我想知道是否有更简单的方法来执行此计算，而不是每次添加新列表时都必须对其进行编辑。我无法更改此流程，因为它支持旧系统并由另一个部门控制。想法？

编辑：Sql Server 2016 SP2。

CREATE TABLE #customers(cust_id int, list_name varchar(10), create_dt date)
INSERT INTO #customers values (1,'Alpha','2019-01-01')
    ,(2,'Alpha','2019-01-01')
    ,(3,'Alpha','2019-01-01')
    ,(4,'Alpha','2019-01-01')
    ,(5,'Alpha','2019-01-01')
    ,(2,'Beta','2019-03-01')
    ,(3,'Beta','2019-03-01')
    ,(4,'Beta','2019-03-01')
    ,(5,'Beta','2019-03-01')
    ,(6,'Beta','2019-03-01')
    ,(7,'Beta','2019-03-01')
    ,(1,'Gamma','2019-05-05')
    ,(6,'Gamma','2019-05-05')
    ,(7,'Gamma','2019-05-05')
    ,(9,'Gamma','2019-05-05')

--Desired Output (long way that needs to be edited every time there is a new list)
SELECT List_Name, 'Starting' Descrip, count(*) Custs FROM #customers WHERE list_name = 'alpha' group by list_name
UNION ALL
SELECT List_Name, 'Add', count(*) FROM #customers a WHERE list_name = 'Beta' AND not exists(SELECT * FROM #customers x WHERE List_Name = 'Alpha' AND a.cust_id = x.cust_id) GROUP BY List_Name
UNION ALL
SELECT 'Beta', 'Remove', -count(*) FROM #customers a WHERE list_name = 'Alpha' AND not exists(SELECT * FROM #customers x WHERE List_Name = 'Beta' AND a.cust_id = x.cust_id) GROUP BY List_Name
UNION ALL
SELECT List_Name, 'Add', count(*) FROM #customers a WHERE list_name = 'Gamma' AND not exists(SELECT * FROM #customers x WHERE List_Name = 'Beta' AND a.cust_id = x.cust_id) GROUP BY List_Name
UNION ALL
SELECT 'Gamma', 'Remove', -count(*) FROM #customers a WHERE list_name = 'Beta' AND not exists(SELECT * FROM #customers x WHERE List_Name = 'Gamma' AND a.cust_id = x.cust_id) GROUP BY List_Name
4

2 回答 2

0

您可以创建一个临时表来生成 id，然后使用这些 id 作为聚合的连接条件，而不是对列表名称进行硬编码。这将允许您在不依赖于知道名称和生成它们的顺序的情况下列出上一个列表。

CREATE TABLE #customers(cust_id int, list_name varchar(10), list_date date)
INSERT INTO #customers
(
    cust_id
    ,list_name
    ,list_date
)
values (1,'Alpha', '01-01-2019')
    ,(2,'Alpha', '01-01-2019')
    ,(3,'Alpha', '01-01-2019')
    ,(4,'Alpha', '01-01-2019')
    ,(5,'Alpha', '01-01-2019')
    ,(2,'Beta', '02-01-2019')
    ,(3,'Beta', '02-01-2019')
    ,(4,'Beta', '02-01-2019')
    ,(5,'Beta', '02-01-2019')
    ,(6,'Beta', '02-01-2019')
    ,(7,'Beta', '02-01-2019')
    ,(1,'Gamma', '03-01-2019')
    ,(6,'Gamma', '03-01-2019')
    ,(7,'Gamma', '03-01-2019')
    ,(9,'Gamma', '03-01-2019')


CREATE TABLE #lists
(
    list_id INT IDENTITY(1,1)
    ,list_name varchar(10)
    ,Starting INT
    ,Added INT
    ,Removed INT
    ,list_date date
)

INSERT INTO #lists
(
    list_name
    ,Starting
    ,Added
    ,Removed
    ,list_date
)
SELECT DISTINCT 
    a.list_name 
    ,Starting = (SELECT COUNT(*) FROM #customers b WHERE b.list_name = a.list_name)
    ,Added = 0
    ,Removed = 0
    ,a.list_date
FROM #customers a
ORDER BY a.list_date ASC

现在使用这个临时表创建另一个具有客户 ID 和列表 ID 的临时表。我这样做是为了不必继续编写连接来链接列表名称和列表 ID。

SELECT c.cust_id
      ,l.list_id
INTO #ListCus
FROM #customers c
INNER JOIN #lists l ON l.list_name = c.list_name

接下来，您可以计算添加和删除的数字。

UPDATE l 
SET l.Added = (SELECT COUNT(*) FROM #ListCus c1 WHERE c1.list_id = l.list_id AND NOT EXISTS (SELECT * FROM #ListCus x WHERE x.list_id = l.list_id-1 AND c1.cust_id = x.cust_id))
    ,l.Removed = (SELECT -COUNT(*) FROM #ListCus c1 WHERE c1.list_id = l.list_id-1 AND NOT EXISTS (SELECT * FROM #ListCus x WHERE x.list_id = l.list_id AND c1.cust_id = x.cust_id)) 
FROM #lists l
WHERE l.list_id > 1 --the first list won't have Added or Removed records

最后，我们通过反透视和过滤掉不需要的描述来格式化数据。

;WITH unpivoted AS
(
    SELECT
        u.list_id
        ,u.Descrip
        ,u.custs
    FROM #lists l 
    UNPIVOT
    (
        custs 
        FOR Descrip IN (Starting, Added, Removed)
    )u
)
,SubResults AS
(
    SELECT u.list_id
          ,u.Descrip
          ,u.custs
    FROM unpivoted u 
    WHERE u.list_id = 1
        AND u.Descrip = 'Starting'

    UNION ALL

    SELECT u.list_id
          ,u.Descrip
          ,u.custs
    FROM unpivoted u 
    WHERE u.list_id <> 1
        AND u.Descrip <> 'Starting'
)
SELECT 
    l.list_name
    ,s.Descrip
    ,s.custs
FROM SubResults s
INNER JOIN #lists l ON l.list_id = s.list_id
ORDER BY s.list_id ASC, s.Descrip
于 2019-05-30T20:12:21.560 回答
0

好的，因此您可以将“列表”存储在公用表表达式中并执行以下操作。它并不性感，但它会给你下面的输出。显然*-1，您可以按照自己的意愿行事。

WITH lists (ListName, ListDate, PreceedingDate)
AS
(
    select distinct list_name, create_dt, PreceedingDate
    from #customers C1
        outer apply (select top 1 PreceedingDate = C3.create_dt
                     from #customers C3
                     where C3.create_dt < C1.create_dt
                     order by C3.create_dt desc) C3
)
select ListID = coalesce(lists.ListDate, CTE2.ListDate)
    , ListName = max(lists.ListName)
    , Added = SUM(IIF(C2.cust_id is null, 1, 0))
    , Removed = SUM(IIF(C1.cust_id is null and C2.cust_id is not null, 1, 0))
    , Remained = SUM(IIF(C2.cust_id = C1.cust_id, 1, 0))
from lists
    inner join #customers C1 on C1.create_dt = lists.ListDate
    full outer join #customers C2 on C1.cust_id = C2.cust_id
                                    and C2.create_dt = lists.PreceedingDate
    --since for removed customers the current List will be NULL
    --we join it back on, which leads to all those COALESCEs
    left join lists CTE2 on CTE2.PreceedingDate = C2.create_dt
where coalesce(lists.ListDate, CTE2.ListDate) is not null
group by coalesce(lists.ListDate, CTE2.ListDate)

结果：

ListID     | ListName | Added | Removed | Remained
-----------|----------|-------|---------|---------
2019-01-01 | Alpha    | 5     | 0       | 0
2019-03-01 | Beta     | 2     | 1       | 4
2019-05-05 | Gamma    | 2     | 4       | 2
于 2019-05-30T20:46:03.847 回答