SQL如何按字段

时间：2018-01-23 17:19:07

标签： sql sql-server limit

我有一个查询，我试图只检索有一个值的1个结果（电话号码字段，可以存储多个电话号码），但是我的结果是复制与供应商关联的每个电话号码的结果。以下是结果集的示例：

enter image description here

从上图中可以看出，NAME1的结果正在重复，因为PHONE字段有3个不同的值。我只想拉1个电话号码（无论哪一个，只要没有空值）。下面是我的sql代码示例，您可以看到注释掉的部分，我试图在C.PHONE的派生查询中捕获MAX值。

SELECT DISTINCT A.NAME1, SUM( A.REMIT_AMT), A.REMIT_VENDOR, (C.PHONE), 
C.PHONE_TYPE
FROM PS_PAYMENT_TBL A, PS_VENDOR B, PS_VENDOR_ADDR_PHN C

WHERE A.PYMNT_DT >= '01-01-2017' 
 AND A.PYMNT_DT <= '12-31-2017' 
 AND A.REMIT_SETID = 'SHARE' 
 AND B.SETID = A.REMIT_SETID 
 AND B.VENDOR_ID = A.REMIT_VENDOR 
 AND B.VENDOR_CLASS <> 'E' 
 AND B.SETID = C.SETID 
 AND B.VENDOR_ID = C.VENDOR_ID
   --AND C.PHONE =
   --(SELECT MAX(C2.PHONE) FROM PS_VENDOR_ADDR_PHN C2)
 AND C.EFFDT = 
    (SELECT MAX(C_ED.EFFDT) FROM PS_VENDOR_ADDR_PHN C_ED 
    WHERE C.SETID = C_ED.SETID 
      AND C.VENDOR_ID = C_ED.VENDOR_ID 
      AND C.ADDRESS_SEQ_NUM = C_ED.ADDRESS_SEQ_NUM 
      AND C_ED.EFFDT <= SUBSTRING(CONVERT(CHAR,GETDATE(),121), 1, 10))

     GROUP BY  A.NAME1,  A.REMIT_VENDOR,  C.PHONE, C.PHONE_TYPE
     ORDER BY 2 DESC

我不相信MS SQL Server像MySQL那样支持LIMIT功能，我可以使用与MS SQL Server兼容的类似内容吗？谢谢！

1/24更新：

SELECT DISTINCT A.NAME1, SUM( A.REMIT_AMT) As TOTAL_SPEND, A.REMIT_VENDOR, 
C.FIRST_PHONE, C.FIRST_PHONE_TYPE
FROM   
PS_PAYMENT_TBL A 

LEFT JOIN (

  SELECT DISTINCT VENDOR_ID,
    FIRST_VALUE(PHONE) OVER (
        PARTITION BY VENDOR_ID
        ORDER BY PHONE DESC
        ROWS UNBOUNDED PRECEDING
    ) AS FIRST_PHONE,
    FIRST_VALUE(PHONE_TYPE) OVER (
        PARTITION BY VENDOR_ID
        ORDER BY PHONE DESC
        ROWS UNBOUNDED PRECEDING
    ) AS FIRST_PHONE_TYPE
  FROM PS_VENDOR_ADDR_PHN C
  WHERE PHONE IS NOT NULL

) C ON A.REMIT_VENDOR = C.VENDOR_ID


, PS_VENDOR B , PS_VENDOR_ADDR_PHN CED

WHERE  A.PYMNT_DT >= '01-01-2017' 
 AND A.PYMNT_DT <= '12-31-2017' 
 AND A.REMIT_SETID = 'SHARE' 
 AND B.SETID = A.REMIT_SETID 
 AND B.VENDOR_ID = A.REMIT_VENDOR 
 AND B.VENDOR_CLASS <> 'E' 
 AND B.SETID = CED.SETID 
 AND B.VENDOR_ID = C.VENDOR_ID

 AND CED.EFFDT = 
    (SELECT MAX(CED.EFFDT) FROM PS_VENDOR_ADDR_PHN CED 
    WHERE CED.SETID = CED.SETID 
      AND CED.VENDOR_ID = CED.VENDOR_ID 
      AND CED.ADDRESS_SEQ_NUM = CED.ADDRESS_SEQ_NUM 
      AND CED.EFFDT <= SUBSTRING(CONVERT(CHAR,GETDATE(),121), 1, 10))

GROUP BY  A.NAME1,  A.REMIT_VENDOR,  C.FIRST_PHONE, C.FIRST_PHONE_TYPE
ORDER BY 2 DESC

1 个答案:

答案 0 :(得分：1)

OP的问题是关于SQL Server中的LIMIT，当实际问题是通过将一个表中的单个记录连接到具有多个记录的其他表（经典XY problem）而引入的重复记录时

让我们尝试确定何时引入重复记录。以下查询没有引入重复记录：

-- Single table, no joins
SELECT A.NAME1, SUM( A.REMIT_AMT), A.REMIT_VENDOR
FROM PS_PAYMENT_TBL A
WHERE 
    A.PYMNT_DT >= '01-01-2017' 
    AND A.PYMNT_DT <= '12-31-2017' 
    AND A.REMIT_SETID = 'SHARE' 
GROUP BY A.NAME1, A.REMIT_VENDOR

据推测，这将返回以下内容：

NAME1                          (No column name) REMIT_VENDOR
------------------------------ ---------------- ------------
Owens & Minor Inc              11828775.730     ERSUT
Besse Medical Supply           11807640.040     BESMED
Medtronic                      11663552.040     MEDINB
Boston Scientific Corporation   8232800.420     SCIMED
Epic Systems Corporation        6841444.300     75128A

但是，由于每个供应商在PS_VENDOR_ADDR_PHN表中可能有多个电话号码，一旦我们JOIN这两个：

SELECT A.NAME1, SUM( A.REMIT_AMT), A.REMIT_VENDOR
FROM PS_PAYMENT_TBL A
INNER JOIN PS_VENDOR_ADDR_PHN C ON A.REMIT_VENDOR = C.VENDOR_ID
WHERE 
    A.PYMNT_DT >= '01-01-2017' 
    AND A.PYMNT_DT <= '12-31-2017' 
    AND A.REMIT_SETID = 'SHARE' 
GROUP BY A.NAME1, A.REMIT_VENDOR

我们将获得重复记录。我们可能不会一开始就注意到它，只要我们在A.REMIT_VENDOR分组;但是由于重复的记录，总数也会变得混乱。

我建议，而不是JOIN在具有重复相关记录的表上，JOIN在子查询上，每个A.REMIT_VENDOR只有一条记录，因此不会引入重复记录。

SELECT A.NAME1, SUM( A.REMIT_AMT), A.REMIT_VENDOR, C.FIRST_PHONE, C.FIRST_PHONE_TYPE
FROM PS_PAYMENT_TBL A
LEFT JOIN (

    -- This subquery returns the first PHONE and PHONE_TYPE, per VENDOR_ID
    -- if the records were ordered by the PHONE in DESC order
    -- FIRST_VALUE is a window function
    SELECT DISTINCT VENDOR_ID,
        FIRST_VALUE(PHONE) OVER (
            PARTITION BY VENDOR_ID
            ORDER BY PHONE DESC
            ROWS UNBOUNDED PRECEDING
        ) AS FIRST_PHONE,
        FIRST_VALUE(PHONE_TYPE) OVER (
            PARTITION BY VENDOR_ID
            ORDER BY PHONE DESC
            ROWS UNBOUNDED PRECEDING
        ) AS FIRST_PHONE_TYPE
    FROM PS_VENDOR_ADDR_PHN
    WHERE PHONE IS NOT NULL

) C ON A.REMIT_VENDOR = C.VENDOR_ID
GROUP BY A.NAME1, A.REMIT_VENDOR

参考文献：

使用JOIN代替WHERE

您的查询不使用JOIN将多组数据相关联，而是将WHERE条件应用于所有数据的笛卡尔积。我自己的感觉是，最好使用JOIN来表达两组数据之间的关系，并使用WHERE专门用于排除数据集中的记录;如果只是因为它可以更容易地调试这些类型的“一个表中的重复记录导致整个结果中的重复记录”问题，那么当您可以看到每个数据集如何与其他数据集相关时。 （请参阅here和here。）

我建议使用JOIN来合并PS_VENDOR表中的数据：

SELECT A.NAME1, SUM( A.REMIT_AMT) AS TOTAL_SPEND, A.REMIT_VENDOR, 
  C.FIRST_PHONE, C.FIRST_PHONE_TYPE
FROM PS_PAYMENT_TBL A 

-- See the INNER JOIN here; it's now easier to understand how PS_PAYMENT_TABLE
-- and PS_VENDOR are related
INNER JOIN PS_VENDOR B
  ON A.REMIT_SETID = B.SETID 
  AND A.REMIT_VENDOR = B.VENDOR_ID

LEFT JOIN (
  SELECT DISTINCT VENDOR_ID,
    FIRST_VALUE(PHONE) OVER (
        PARTITION BY VENDOR_ID
        ORDER BY PHONE DESC
        ROWS UNBOUNDED PRECEDING
    ) AS FIRST_PHONE,
    FIRST_VALUE(PHONE_TYPE) OVER (
        PARTITION BY VENDOR_ID
        ORDER BY PHONE DESC
        ROWS UNBOUNDED PRECEDING
    ) AS FIRST_PHONE_TYPE
  FROM PS_VENDOR_ADDR_PHN C
  WHERE PHONE IS NOT NULL
) C ON A.REMIT_VENDOR = C.VENDOR_ID,

WHERE
   A.PYMNT_DT >= '01-01-2017' 
  AND A.PYMNT_DT <= '12-31-2017' 
  AND A.REMIT_SETID = 'SHARE'
   -- with the JOIN, we can apply filtering conditions on data from the B table
  AND B.VENDOR_CLASS <> 'E'

GROUP BY A.NAME1, A.REMIT_VENDOR
ORDER BY 2 DESC

合并EEFDT字段

（仍在审理中）

原始答案

SQL Server的相应语法是：

AND C.PHONE =
   (SELECT TOP 1 C2.PHONE FROM PS_VENDOR_ADDR_PHN C2)

这将返回单个任意PHONE。要返回最大PHONE，请对子查询记录进行排序：

AND C.PHONE =
   (SELECT TOP 1 C2.PHONE FROM PS_VENDOR_ADDR_PHN C2 ORDER BY C2.PHONE DESC)

参考 - TOP clause

相关问题
最新问题