SQL:3个自联接然后将它们连接在一起

时间:2009-04-08 16:00:58

标签: sql sql-server sql-server-2005 join

我有2个表以特定方式加入。我认为我的查询是正确的,但不确定。

select t1.userID, t3.Answer Unit, t5.Answer Demo  
FROM    
    table1 t1  
    inner join (select * from table2) t3 ON t1.userID = t3.userID   
    inner join (select * from table2) t5 ON t1.userID = t5.userID  
where  
    NOT EXISTS (SELECT * FROM table1 t2 WHERE t2.userID = t1.userID AND t2.date > t1.date)  
    and NOT EXISTS (SELECT * FROM table2 t4 WHERE t4.userID = t3.userID and t4.counter > t3.counter)  
    and NOT EXISTS (SELECT * FROM table2 t6 WHERE t6.userID = t5.userID and t6.counter > t5.counter)  
    and t1.date_submitted >'1/1/2009'  
    and t3.question = Unit  
    and t5.question = Demo  
order by  
    t1.userID  

从table1我想要不同的userID,其中date> 1/1/2009

table1      
userID    Date
1         1/2/2009  
1         1/2/2009  
2         1/2/2009  
3         1/2/2009  
4         1/1/2008  

所以我想从table1得到的结果应该是:

userID  
1  
2  
3  

然后我想在tableID上使用table2加入它,如下所示:

table2 
userID    question   answer   counter
1         Unit       A        1
1         Demo       x        1
1         Prod       100      1
2         Unit       B        1
2         Demo       Y        1
3         Prod       100      1
4         Unit       A        1
1         Unit       B        2
1         Demo       x        2
1         Prod       100      2
2         Unit       B        2
2         Demo       Z        2
3         Prod       100      2
4         Unit       A        2

我希望将table1与table2结合使用此结果:

userID    Unit    Demo
1         B       X
2         B       Z

换句话说,
从table2中选择不同的userID,其中question =最高计数器的单位
然后
从table2中选择不同的userID,其中question =最高计数器的Demo。

我认为我所做的是3次自我加入然后将这3个加入。

你认为这是对的吗?

1 个答案:

答案 0 :(得分:3)

SELECT  du.userID, unit.answer, demo.answer
FROM    (
        SELECT  DISTINCT userID
        FROM    table1
        WHERE   date > '1/1/2009'
        ) du
LEFT JOIN
        table2 unit
ON      (userID, question, counter) IN
        (
        SELECT  du.userID, 'Unit', MAX(counter)
        FROM    table2 td
        WHERE   userID = du.userID
                AND question = 'Unit'
        )
LEFT JOIN
        table2 demo
ON      (userID, question, counter) IN
        (
        SELECT  du.userID, 'Demo', MAX(counter)
        FROM    table2 td
        WHERE   userID = du.userID
                AND question = 'Demo'
        )

table2 (userID, question, counter)上建立索引将大大改善此查询。

由于您提到SQL Server 2005,以下内容将更容易,更有效:

SELECT  du.userID,
        (
        SELECT  TOP 1 answer
        FROM    table2 ti
        WHERE   ti.user = du.userID
                AND ti.question = 'Unit'
        ORDER BY
                counter DESC
        ) AS unit_answer,
        (
        SELECT  TOP 1 answer
        FROM    table2 ti
        WHERE   ti.user = du.userID
                AND ti.question = 'Demo'
        ORDER BY
                counter DESC
        ) AS demo_answer
FROM    (
        SELECT  DISTINCT userID
        WHERE   date > '1/1/2009'
        FROM    table1
        ) du

聚合:

SELECT  answer, COUNT(*)
FROM    (
        SELECT  DISTINCT userID
        FROM    table1
        WHERE   date > '1/1/2009'
        ) du
JOIN    table2 t2
ON      t2.userID = du.userID
        AND t2.question = 'Unit'
GROUP BY
        answer