使用来自第二个表的不同值的 T-SQL 连接查询

Question

我知道这个问题听起来像是重复的问题，但我已经解决了我能找到的所有问题；虽然它仍然有可能与我可能错过的问题重复。

我有什么表面价值似乎是微不足道的要求，但无论我如何编写它，总有一些警告是行不通的。我试过GROUP、DISTINCT、JOIN、聚合函数等。

场景： PRIMARYTABLE 包含一组活动，SECONDARYTABLE 包含活动的运行日期。每个广告系列可以有多次运行，我为每次运行添加了SUBKEY。

要求： 我需要能够将最近运行的广告系列放入列表中，以便用户可以更轻松地从运行最频繁的广告系列中进行选择。

PRIMARYTABLE
KEYCOLUMN   INFOCOLUMN
100000      Test 1
100001      Test Campaign
100002      Test Image 2
100003      Test Img
100004      Image Test
100005      Test
100006      Test Image 3
100007      Test Image 4
100008      Test Image 5
100009      Image Comparison Test 2
100010      Testing
100011      Test Fields
100012      Test 5
100013      test

SECONDARYTABLE
KEYCOLUMN   SUBKEY  DATECOLUMN
100000      100000  2017-06-02 04:09:57.593
100001      100001  2017-06-19 12:09:54.093
100001      100002  2017-06-27 10:51:14.140
100004      100003  2017-06-27 12:33:47.747
100006      100004  2017-06-28 10:29:53.387
100007      100005  2017-06-28 10:36:23.710
100008      100006  2017-06-29 22:31:03.790
100009      100007  2017-06-29 23:07:52.870
100009      100010  2017-10-04 16:05:40.583
100009      100011  2017-10-04 16:09:55.470
100011      100008  2017-09-08 14:02:28.017
100012      100009  2017-09-11 16:17:23.870
100013      100012  2017-11-07 16:55:55.403
100013      100013  2017-11-08 15:37:16.430

以下是我所追求的或多或少的想法。

SELECT DISTINCT( a.[INFOCOLUMN] )
FROM [PRIMARYTABLE] a
INNER JOIN [SECONDARYTABLE] b ON ( a.[KEYCOLUMN] = b.[KEYCOLUMN] )
ORDER BY a.[DATECOLUMN]

希望有一个荷马辛普森“Doh！”当我看到它应该如何完成时。

非常感谢。

Answer 1

1. the most recently run campaigns >> 使用 row_number() over(.. order by ... DESC)
2. that get run the most frequent >> 使用 count(*) over(partition by ..)

使用窗口函数row_number() over() 和count() over() 可以按“最近”的数据行进行选择，并按“最频繁”进行排序。请注意，日期的 DESCending 顺序会导致 "recent" = 1。

select
       p.*, s.*
from PRIMARYTABLE p
inner join (
      select KEYCOLUMN, SUBKEY, DATECOLUMN
           , row_number() over(partition by KEYCOLUMN order by DATECOLUMN DESC) recent
           , count(*) over(partition by KEYCOLUMN) frequency
      from SECONDARYTABLE
      ) s on p.KEYCOLUMN = s.KEYCOLUMN  and s.recent = 1
order by s.frequency DESC, p.INFOCOLUMN

Answer 2

你可以试试这个：

DECLARE @PRIMARYTABLE TABLE
(
    [KEYCOLUMN] INT 
   ,[INFOCOLUMN] VARCHAR(24)
);

DECLARE @SECONDARYTABLE TABLE
(
    [KEYCOLUMN] INT 
   ,[SUBKEY] INT
   ,[DATECOLUMN] DATETIME2
);

INSERT INTO @PRIMARYTABLE ([KEYCOLUMN], [INFOCOLUMN])
VALUES (100000, 'Test 1')
      ,(100001, 'Test Campaign')
      ,(100002, 'Test Image 2')
      ,(100003, 'Test Img')
      ,(100004, 'Image Test')
      ,(100005, 'Test')
      ,(100006, 'Test Image 3')
      ,(100007, 'Test Image 4')
      ,(100008, 'Test Image 5')
      ,(100009, 'Image Comparison Test 2')
      ,(100010, 'Testing')
      ,(100011, 'Test Fields')
      ,(100012, 'Test 5')
      ,(100013, 'test');

INSERT INTO @SECONDARYTABLE ([KEYCOLUMN], [SUBKEY], [DATECOLUMN])
VALUES (100000, 100000, '2017-06-02 04:09:57.593')
      ,(100001, 100001, '2017-06-19 12:09:54.093')
      ,(100001, 100002, '2017-06-27 10:51:14.140')
      ,(100004, 100003, '2017-06-27 12:33:47.747')
      ,(100006, 100004, '2017-06-28 10:29:53.387')
      ,(100007, 100005, '2017-06-28 10:36:23.710')
      ,(100008, 100006, '2017-06-29 22:31:03.790')
      ,(100009, 100007, '2017-06-29 23:07:52.870')
      ,(100009, 100010, '2017-10-04 16:05:40.583')
      ,(100009, 100011, '2017-10-04 16:09:55.470')
      ,(100011, 100008, '2017-09-08 14:02:28.017')
      ,(100012, 100009, '2017-09-11 16:17:23.870')
      ,(100013, 100012, '2017-11-07 16:55:55.403')
      ,(100013, 100013, '2017-11-08 15:37:16.430');


SELECT a.[INFOCOLUMN] 
      ,b.[DATECOLUMN]
FROM @PRIMARYTABLE A
CROSS APPLY
(
    SELECT TOP 1 [DATECOLUMN]
    FROM @SECONDARYTABLE  B
    WHERE A.[KEYCOLUMN] = B.[KEYCOLUMN]
    ORDER BY [DATECOLUMN] DESC
) b;

它将为您提供每个广告系列的最后一次执行。然后您可以按日期或 ORDER BY 过滤并从最终查询中获取 TOP N。

或者你可以使用ROW_NUMBER:

WITH DataSource AS
(
    SELECT A.[INFOCOLUMN]
          ,B.[DATECOLUMN]
          ,ROW_NUMBER() OVER (PARTITION BY A.[KEYCOLUMN] ORDER BY B.[KEYCOLUMN]) AS [RowID]
    FROM @PRIMARYTABLE A
    INNER JOIN @SECONDARYTABLE B
        ON A.[KEYCOLUMN] = B.[KEYCOLUMN]
)
SELECT [INFOCOLUMN]
      ,[DATECOLUMN]
FROM DataSource
WHERE [RowID] = 1;

Answer 3

试试这个，它会以最频繁的使用顺序返回广告系列列表。注意从不运行的广告系列不会出现在您的列表中。在这种情况下，您将进行左连接

SELECT a.[INFOCOLUMN] 
FROM   [PRIMARYTABLE] a 
 /* left */ JOIN [SECONDARYTABLE] b ON a.[KEYCOLUMN] = b.[KEYCOLUMN] 
group BY a.[infocolumn]
order by max(datecolumn) desc

这是我用来测试它的存根

select 10000 id,'Campain A' cname into #a1 union all
select 10002,'Campain B' union all
select 10004,'Campain C' union all
select 10009,'Campain E' 

select 10000 id,'20170101' thedate into #a2 union all
select 10000,'20170102' union all
select 10009,'20170103' union all
select 10002,'20170104' union all
select 10004,'20170105' union all
select 10000,'20170201' union all
select 10000,'20170302' union all
select 10009,'20170403' union all
select 10002,'20170104' union all
select 10004,'20170205' union all
select 10000,'20170101' union all
select 10004,'20170302' union all
select 10000,'20170103' union all
select 10002,'20170404' union all
select 10002,'20170105' 

select #a1.cname
 from #a1 join #a2 on #a1.id = #a2.id 
 group by #a1.cname
 order by max(thedate) desc