Question

我创建了以下视图“user_details_merged”：

SELECT DISTINCT
coalesce(own.user_name, join_user_name.user_name) AS user_name,
coalesce(own.email, join_mail.email) AS email,
coalesce(own.first_name, join_name.first_name) AS first_name,
coalesce(own.last_name, join_name.last_name) AS last_name
FROM
user_details AS own

LEFT JOIN user_details AS join_user_name ON 
    own.user_name IS NULL AND (
    (join_user_name.email = own.email AND own.email IS NOT NULL) 
    OR (join_user_name.first_name = own.first_name AND join_user_name.last_name = own.last_name 
    AND own.first_name IS NOT NULL AND own.last_name IS NOT NULL))


LEFT JOIN user_details AS join_mail ON 
    own.email IS NULL AND (
    (join_mail.user_name = own.user_name AND own.user_name IS NOT NULL) 
    OR (join_mail.first_name = own.first_name AND join_mail.last_name = own.last_name 
    AND own.first_name IS NOT NULL AND own.last_name IS NOT NULL))

LEFT JOIN user_details AS join_name ON 
    own.first_name IS NULL AND own.last_name IS NULL AND (
    (join_name.email = own.email AND own.email IS NOT NULL) 
    OR (join_name.user_name = own.user_name AND own.user_name IS NOT NULL))

ORDER BY user_name ASC,email ASC, first_name ASC, last_name ASC

这将合并我的列：

user_name | email | first_name | last_name
a             b       NULL         NULL
NULL          b        c            d
a            NULL      e            f
NULL          x        y            z

到

user_name | email | first_name | last_name
a             b       NULL         NULL
NULL          b        c            d
a            NULL      e            f
NULL          x        y            z

a             b        c            d
a             b        e            f

我想要的是：

user_name | email | first_name | last_name
NULL          x        y            z
a             b        c            d
a             b        e            f

当没有包含ROWS的NULL时，如果有一行具有相同的数据，但是当没有其他行包含更多信息时，仍然保持/ NULL x y z /。

这里的第二个视图完全符合我的需要：

SELECT DISTINCT a.user_name,a.email,a.first_name,a.last_name FROM
user_details_merged a
LEFT JOIN user_details_merged b
ON
(
    (
    a.user_name IS NOT NULL OR 
    NOT EXISTS (SELECT user_name FROM user_details_merged b WHERE b.user_name IS NOT NULL AND 
        b.email=ISNULL(a.email,b.email) AND 
        b.first_name=isnull(a.first_name,b.first_name) AND
        b.last_name=isnull(a.last_name,b.last_name))
    )

    AND

    (
    a.email IS NOT NULL OR 
    NOT EXISTS (SELECT email FROM user_details_merged b WHERE b.email IS NOT NULL AND 
        b.user_name=ISNULL(a.user_name,b.user_name) AND 
        b.first_name=isnull(a.first_name,b.first_name) AND
        b.last_name=isnull(a.last_name,b.last_name))
    )

    AND

    (
    (a.first_name IS NOT NULL AND a.last_name IS NOT NULL) OR 
    NOT EXISTS (SELECT email FROM user_details_merged b WHERE b.email IS NOT NULL AND 
        b.user_name=ISNULL(a.user_name,b.user_name) AND 
        b.email=ISNULL(a.email,b.email))
        -- AND b.first_name=isnull(a.first_name,b.first_name) AND b.last_name=isnull(a.last_name,b.last_name))
    )

    AND NOT (a.first_name = b.first_name AND a.last_name = b.last_name AND a.email = b.email AND a.user_name = b.user_name)

)

WHERE coalesce(b.user_name,b.email,b.first_name,b.last_name) IS NOT NULL

主要问题是user_details视图（数据来自哪里）由许多不同表的联合组成。一些只包含用户名和＆amp;电子邮件和一些只有电子邮件和名字/姓氏等。这就是为什么没有唯一的密钥，因为UNIONS我无法索引视图。这使得无法在一小时内执行最后一个视图。我目前的解决方法是一个过程，它将user_details_merged视图的数据存储在临时表中，让上面的第二个视图使用该表中的数据。这样我可以将8000行的执行时间减少到7秒。

还有其他建议吗？

非常感谢你;）

Answer 1

哎哟！这是一个混乱的数据模型。最好的解决方案是修复数据模型以防止需要这样复杂的查询。但是，随着应用程序的依赖性，这往往更加复杂，所以我假设已经受理了。

我使用了4行样本来提出替代解决方案。
然后我添加了几行，只有First和Last名称值，这些行在上面的查询中暴露了错过的场景。
我还将相同的6行复制到超过12K行，因为数据模型似乎支持这种情况。这最终导致上述查询运行超过2小时，然后我终于放弃并停止了它。
我针对我的解决方案运行了12K行，并且在不到一秒的时间内返回了预期的结果。

所以，没有进一步的说明：

-- =================================================================================
-- BEGIN: SETUP TEST DATA
-- =================================================================================
SET NOCOUNT ON 

IF OBJECT_ID('user_details', 'U') IS NOT NULL DROP TABLE user_details;
GO

CREATE TABLE dbo.user_details (
    user_name   char(1) NULL,
    email       char(1) NULL,
    first_name  char(1) NULL,
    last_name   char(1) NULL
)
GO

INSERT dbo.user_details
SELECT * 
  FROM (
        SELECT * FROM dbo.user_details WHERE 1=2
        UNION ALL SELECT 'a',   'b',    NULL,   NULL
        UNION ALL SELECT NULL,  'b',    'c',    'd'
        UNION ALL SELECT 'a',   NULL,   'e',    'f'
        UNION ALL SELECT NULL,  'x',    'y',    'z'
        UNION ALL SELECT NULL,  NULL,   'y',    'z'
        UNION ALL SELECT NULL,  NULL,   'a',    'z'
       ) A
GO

--/*
-- TURN 6 ROWS INTO OVER 12K ROWS TO TEST PERFORMANCE
DECLARE @count int; SELECT @count = 0
WHILE @count < 11
  BEGIN 
    INSERT user_details
    SELECT * 
      FROM user_details

    SELECT @count = @count + 1
END
--*/
-- =================================================================================
-- END: SETUP TEST DATA
-- =================================================================================


-- =================================================================================
-- BEGIN: NEW SOLUTION FINAL: <1sec on 12288 rows
-- =================================================================================
IF OBJECT_ID('tempdb..#useremail', 'U') IS NOT NULL DROP TABLE #useremail;
IF OBJECT_ID('tempdb..#email', 'U') IS NOT NULL DROP TABLE #email;
IF OBJECT_ID('tempdb..#user', 'U') IS NOT NULL DROP TABLE #user;
IF OBJECT_ID('tempdb..#name', 'U') IS NOT NULL DROP TABLE #name;


-- GET YOUR UNIQUE user_name AND email KEY
SELECT DISTINCT A.user_name, A.email
  INTO #useremail
  FROM user_details A


-- GET YOUR UNIQUE email VALUES
SELECT DISTINCT A.email, A.first_name, A.last_Name
  INTO #email
  FROM user_details A
 WHERE A.email IS NOT NULL


-- GET YOUR UNIQUE user_name VALUES
SELECT DISTINCT A.user_name, A.first_name, A.last_Name
  INTO #user
  FROM user_details A
 WHERE A.user_name IS NOT NULL


-- GET YOUR UNIQUE first_name AND last_Name VALUES NOT PART OF THE KEY
SELECT DISTINCT A.first_name, A.last_Name
  INTO #name
  FROM user_details A
 WHERE A.first_name IS NOT NULL
   AND A.last_Name IS NOT NULL
   AND A.user_name IS NULL
   AND A.email IS NULL


-- CLEAN UP YOUR UNIQUE user_name AND email KEY
DELETE A
-- SELECT *
  FROM #useremail A
  JOIN (
        SELECT *
          FROM #useremail
         WHERE user_name IS NOT NULL
           AND email IS NOT NULL
       ) B
    ON (A.user_name = B.user_name AND A.email     IS NULL)
    OR (A.email     = B.email     AND A.user_name IS NULL)


-- CLEAN UP YOUR UNIQUE email VALUES
DELETE A
-- SELECT *
  FROM #email A
  JOIN (
        SELECT *
          FROM #email
         WHERE first_name IS NOT NULL
           AND last_Name IS NOT NULL
       ) B
    ON A.email = B.email
   AND A.first_name IS NULL
   AND A.last_name IS NULL


-- CLEAN UP YOUR UNIQUE user_name VALUES
DELETE A
-- SELECT *
  FROM #user A
  JOIN (
        SELECT *
          FROM #user
         WHERE first_name IS NOT NULL
           AND last_Name IS NOT NULL
       ) B
    ON A.user_name = B.user_name
   AND A.first_name IS NULL
   AND A.last_name IS NULL


-- CLEAN UP YOUR UNIQUE #name VALUES
DELETE A
-- SELECT *
  FROM #name A
  JOIN #user B
    ON A.first_name = B.first_name
   AND A.last_name  = B.last_name

DELETE A
-- SELECT *
  FROM #name A
  JOIN #email B
    ON A.first_name = B.first_name
   AND A.last_name  = B.last_name


-- GET YOUR DATA
SELECT A.user_name
      ,A.email
      ,U.first_name
      ,U.last_name
      --,*
  FROM #useremail A
  JOIN #user U
    ON A.user_name = U.user_name
 UNION 
SELECT A.user_name
      ,A.email
      ,E.first_name
      ,E.last_name
      --,*
  FROM #useremail A
  JOIN #email E
    ON A.email = E.email
 UNION 
SELECT NULL as [user_name]
      ,NULL as [email]
      ,N.first_name
      ,N.last_name
      --,*
  FROM #name N
-- =================================================================================
-- END: NEW SOLUTION FINAL
-- =================================================================================

Answer 2

- 如果我这次说得对，你可以通过以下方式解决：

 SELECT ISNULL(A.USER_NAME, B.USER_NAME), A.EMAIL, A.FIRST_NAME, A.LAST_NAME FROM
 user_details A CROSS JOIN user_details B
 WHERE A.EMAIL IS NOT NULL
 AND A.FIRST_NAME IS NOT NULL
 AND A.LAST_NAME IS NOT NULL
 GROUP BY ISNULL(A.USER_NAME, B.USER_NAME), A.EMAIL, A.FIRST_NAME, A.LAST_NAME

Answer 3

尝试对user_details使用外连接。

消除合并连接结果集中的NULL字段

3 个答案: