为啥某些 Where 子句被忽略？

Posted 2023-02-16

【中文标题】为啥某些 Where 子句被忽略？

【英文标题】：Why is some of the Where clause being ignored?为什么某些 Where 子句被忽略？

【发布时间】：2021-01-08 23:11:32

【问题描述】：

我有这张桌子：

CREATE TABLE [dbo].[Handshake]
(
    [Report Year] [varchar](100) NULL,
    [Status] [varchar](100) NULL,
    [Update Time] [datetime] NULL,
    [Process Time] [datetime] NULL
) ON [PRIMARY]
GO

一些样本数据：

Report Year Status       Update Time                Process Time
----------------------------------------------------------------
Test        Loading      2020-09-22 12:53:41.417    NULL
2020 1+7    Loaded       2020-09-22 12:46:41.417    NULL
2020 2+7    Loaded       2020-09-21 21:18:25.130    NULL

我有以下程序：

DECLARE
    @ReportYearCmd VARCHAR(1000),
    @CosCountCmd VARCHAR(1000),
    @FranchiseCountCmd AS VARCHAR(1000),
    @ProductCountCmd AS VARCHAR(1000);

WITH Validations AS 
(
    SELECT TOP 1 * 
    FROM [Handshake] 
    WHERE [Status] <> 'Loading' 
      AND [Update Time] = (SELECT MAX([Update Time]) FROM Handshake)
)
UPDATE Validations 
SET @ReportYearCmd = CASE 
                        WHEN Report_Year_Count = 0 
                           THEN NULL 
                           ELSE 'SELECT DISTINCT [Report Year] AS [Report Year] FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [Report Year] WHERE [Report Year].[Report Year] = [Fact].[Report Year] );'
                     END,
    @CosCountCmd = CASE 
                      WHEN COS_Count = 0 
                         THEN NULL 
                         ELSE 'SELECT DISTINCT [Country Code] AS [COS - Country Code] FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [COS] WHERE [COS].[Country Code] = [Fact].[Country Code] );' 
                   END,
    @FranchiseCountCmd = CASE WHEN  Franchise_Count = 0 THEN NULL 
        ELSE 'SELECT DISTINCT [Style Code] AS [Franchise - Style Code] FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [Franchise] WHERE [Franchise].[Style Code] = [Fact].[Style Code] );' 
    END,
    @ProductCountCmd = CASE WHEN Product_Count = 0 THEN NULL 
        ELSE 'SELECT DISTINCT [Style Code] AS [Product - Style Code] FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [Product] WHERE [Product].[Style Code] = [Fact].[Style Code] );' 
    END,
    [Status] = CASE
        WHEN ( Report_Year_Count = 0 AND COS_Count = 0 AND Franchise_Count = 0 AND Product_Count = 0 ) THEN 'Good'
        ELSE 'Rejects'
    END
FROM 
    [Validations] 
OUTER APPLY 
    (SELECT 
         ISNULL((SELECT COUNT(*) FROM [Fact] 
                 WHERE NOT EXISTS (SELECT * FROM [Report Year] WHERE [Report Year].[Report Year] = [Fact].[Report Year] ) ), 0 ) AS [Report_Year_Count],
        ISNULL( ( SELECT COUNT(*) FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [COS] WHERE [COS].[Country Code] = [Fact].[Country Code] ) ), 0 ) AS [COS_Count],
        ISNULL( ( SELECT COUNT(*) FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [Franchise] WHERE [Franchise].[Style Code] = [Fact].[Style Code] ) ), 0 ) AS [Franchise_Count],
        ISNULL( ( SELECT COUNT(*) FROM [Fact] WHERE NOT EXISTS ( SELECT * FROM [Product] WHERE [Product].[Style Code] = [Fact].[Style Code] ) ), 0 ) AS [Product_Count]

) AS [ValidationCounts];

只有当我在 Validations SELECT 语句中不包含此部分时，该过程才有效：...WHERE [Status] <> 'Loading'...，因此会选择最近的更新时间记录，例如

Test        Loading      2020-09-22 12:53:41.417    NULL

当我包含...WHERE [Status] <> 'Loading'... 时，什么都没有发生！

期望：查询应该获取这条记录：

2020 1+7    Loaded       2020-09-22 12:46:41.417    NULL

但它没有……为什么？

【问题讨论】：

试试这个...WHERE [Status] != N'Loading'... 或...WHERE [Status] not in (N'Loading')...

@Amirhossein nope 没用

【参考方案1】：

如果你想要最新的未加载状态，则不需要子查询。您可以在带有where 子句的非加载行上使用过滤器，然后order by 降序日期，并仅保留top (1) 行：

WITH validations AS (
    SELECT TOP (1) * 
    FROM [Handshake] 
    WHERE [Status] <> 'Loading' 
    ORDER BY [Update Time] DESC
)
UPDATE validations ...

【讨论】：

这个完全符合我的预期，并且不更新旧记录:)

【参考方案2】：

SELECT MAX( [Update Time] ) FROM Handshake 的结果是 2020-09-22 12:53:41.417

现在您的查询是

SELECT TOP 1 * FROM [Handshake] WHERE [Status] <> 'Loading' AND [Update Time] = '2020-09-22 12:53:41.417'

当您评估上述查询时，您会看到在更新时间 2020-09-22 12:53:41.417 时没有状态加载的记录

正确的查询应该是

SELECT TOP 1 * FROM [Handshake] h 
WHERE [Status] <> 'Loading' AND 
  [Update Time] = ( 
    SELECT MAX( [Update Time] ) 
    FROM Handshake hi where hi.Status = h.Status)

【讨论】：

哦，等一下，如果我一遍又一遍地执行它，它现在也会更新旧记录......它应该只更新最新的非加载状态

但最新的非加载记录在您​​更新时会发生变化。你还有另一个顾虑。您正在更改状态，这实际上是预期的。这是一个设计问题。

是对“拒绝”或“不拒绝”的最新非加载更改。但不应该碰旧的记录。应该只更新最近的记录，因此为什么我要以SELECT MAX( [Update Time] ) 开头

那么您的查询就可以了。在给定时间没有处于非加载状态的记录。

但是当我再次运行查询时这条记录得到了更新，它不应该是....2020 2+7 Loaded 2020-09-21 21:18:25.130 NULL