将“UNION”与“IF EXISTS”一起使用

Posted 2023-05-08

【中文标题】将“UNION”与“IF EXISTS”一起使用

【英文标题】：Using "UNION" with "IF EXISTS"

【发布时间】：2015-08-03 19:51:01

【问题描述】：

我在 SQL Server 2008 R2 上遇到以下问题，我正在尝试对一组 UNION 使用 IF EXISTS 子句，但出现错误。

如果我单独运行查询，它们运行时没有错误，但如果我尝试使用 UNION 运行它们，它给了我错误：

消息 156，第 15 级，状态 1，第 9 行 关键字“IF”附近的语法不正确。

SELECT
    '1.5.2- Customers' as INPUT,
    count (distinct ID) as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
FROM
    [BASE]
WHERE
    BASE.CAT_CUSTO = 'EMPLO' and
    BASE.ESTAT = 'EX'
group by
    PRO_DATE

UNION 

IF  EXISTS (
    SELECT
        '1.6- OTHER CUSTOMERS' as INPUT,
        count (distinct ID)  as Z,
        PRO_DATE,
        Month(PRO_DATE) as P_MONTH,
        Year(PRO_DATE) as P_YEAR
    FROM
        [BASE]
    WHERE
        BASE.CAT_CUSTO <> 'EMPLO'
    group by
        PRO_DATE
)  begin
SELECT
        '1.6- OTHER CUSTOMERS' as INPUT,
        count (distinct ID)  as Z,
        PRO_DATE,
        Month(PRO_DATE) as P_MONTH,
        Year(PRO_DATE) as P_YEAR
    FROM
        [BASE]
    WHERE
        BASE.CAT_CUSTO <> 'EMPLO'
    group by
        PRO_DATE

 end
ELSE
    SELECT TOP 1
        '1.6- OTHER CUSTOMERS' as INPUT,
        0 as Z,
        PRO_DATE,
        Month(PRO_DATE) as P_MONTH,
        Year(PRO_DATE) as P_YEAR
    FROM
        [BASE]

【问题讨论】：

您混淆了 T-SQL 控制流代码和 SQL 查询。不是一个不寻常的混乱，但你最好描述你想要做什么（即样本数据和期望的结果）而不是显示显然无法工作的代码。

Top 1 without order by 只是从表中随机选择一条记录。

@Magnus 是的，这也让我感到困惑。

带有“前 1”的最后一部分仅获得 1 条记录，因为如果我有 100 条，我将得到 100 条相同的结果（表中每条记录的结果）

@ThatguyfromIT 你可以接受随机的PRO_DATE吗？

【参考方案1】：

你说

q1

union 

if exists(q2)

  q3

else

  q4

不起作用，但所有查询都起作用。我怀疑它查看查询，但如果是这种情况，那么只需将其更改为

if exists(q2)

  q1

  union 

  q3

else

  q1

  union 

  q4

【讨论】：

nono，就像 q1 union if exists(q2) q2 else q3

@ThatguyfromIT - 好的...我在第二个参考中调用 q2 q3 的事实与解决方案的逻辑无关。

【参考方案2】：

您可以使用临时表：

SELECT
    '1.6- OTHER CUSTOMERS' as INPUT,
    count (distinct ID)  as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
  INTO #temp
FROM
    [BASE]
WHERE
    BASE.CAT_CUSTO <> 'EMPLO'
group by
    PRO_DATE

然后

IF SELECT COUNT(*) FROM #temp > 0
BEGIN
SELECT
    '1.5.2- Customers' as INPUT,
    count (distinct ID) as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
FROM
    [BASE]
WHERE
    BASE.CAT_CUSTO = 'EMPLO' and
    BASE.ESTAT = 'EX'
group by
    PRO_DATE

UNION 

SELECT
    '1.6- OTHER CUSTOMERS' as INPUT,
    count (distinct ID)  as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
FROM
    [BASE]
WHERE
    BASE.CAT_CUSTO <> 'EMPLO'
group by
    PRO_DATE

 end

ELSE

SELECT
    '1.5.2- Customers' as INPUT,
    count (distinct ID) as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
FROM
    [BASE]
WHERE
    BASE.CAT_CUSTO = 'EMPLO' and
    BASE.ESTAT = 'EX'
group by
    PRO_DATE

UNION 

SELECT TOP 1
    '1.6- OTHER CUSTOMERS' as INPUT,
    0 as Z,
    PRO_DATE,
    Month(PRO_DATE) as P_MONTH,
    Year(PRO_DATE) as P_YEAR
FROM
    [BASE]

【讨论】：

听起来不错，让我试一试，因为我希望编辑 ssis 以分开执行这些步骤

【参考方案3】：

DECLARE @chk NVARCHAR(1)='B';
/** change @chk as 'C' to see the result**/
WITH T1 AS(SELECT * FROM(VALUES('A',1))a(F1,F2)
),T2 AS(SELECT * FROM(VALUES('B',2))a(F1,F2)
),T3 AS(SELECT * FROM(VALUES('C',3))a(F1,F2)
),CHK AS(SELECT CASE WHEN EXISTS(SELECT * FROM T2 WHERE F1=@chk) THEN 'B' ELSE 'C' END[chk_f]
) SELECT * FROM T1 UNION SELECT T2.* FROM T2,CHK WHERE chk_f='B' UNION SELECT T3.* FROM T3,CHK WHERE chk_f='C'

【讨论】：

请不要只发布代码作为答案，还要解释您的代码的作用以及它如何解决问题的问题。带有解释的答案通常更有帮助、质量更好，并且更有可能吸引投票