挑出 EVENTDATA 的特定部分用于 DDL 触发器?

Single out specific parts of EVENTDATA for use in a DDL Trigger?

当前触发器如下:

CREATE TRIGGER TestTrigger
ON DATABASE
FOR ALTER_TABLE
AS
BEGIN
SET NOCOUNT ON;
    DECLARE @FULL_STATEMENT SYSNAME
    SELECT @FULL_STATEMENT = EVENTDATA().value('(/EVENT_INSTANCE/TSQLCommand/CommandText)[1]','nvarchar(max)')
    print('The following SQL statement was executed:')
    print('')
    print (@FULL_STATEMENT)
    print('')
    print('Next step is to work out how to single out the added column name')
    ROLLBACK
END

然后我运行下面SQL声明:

USE [a_Database]
ALTER TABLE dbo.TABLE
ADD TestColumn varchar(50)

它输出以下内容:

The following SQL statement was executed:

ALTER TABLE dbo.TABLE

ADD TestColumn varchar(50)

Next step is to work out how to single out the added column name

Msg 3609, Level 16, State 2, Line 2

The transaction ended in the trigger. The batch has been aborted.

最终目标是挑出"TestColumn" 以确保它不包含可能破坏内部程序的特定字符。 "TestColumn" 将是一个有效名称,但 "Test.Column" 则不是,例如。

将它挑出来并将其传递给新变量(“@ColumnName”?)以检查不需要的字符的最佳方法是什么?

您需要记住,一条 alter 语句可以添加多个列。如果您查看为 alter column 语句生成的 XML,您会得到一个更好的主意:

ALTER TABLE dbo.EventTest ADD NewColumn1 INT, NewColumn2 INT;

----------------------------
<EVENT_INSTANCE>
  <EventType>ALTER_TABLE</EventType>
  <PostTime>2015-06-30T14:28:30.790</PostTime>
  <SPID>67</SPID>
  <ServerName>XXXXXX</ServerName>
  <LoginName>XXXXXX</LoginName>
  <UserName>dbo</UserName>
  <DatabaseName>XXXXXX</DatabaseName>
  <SchemaName>dbo</SchemaName>
  <ObjectName>EventTest</ObjectName>
  <ObjectType>TABLE</ObjectType>
  <AlterTableActionList>
    <Create>
      <Columns>
        <Name>NewColumn1</Name>
        <Name>NewColumn2</Name>
      </Columns>
    </Create>
  </AlterTableActionList>
  <TSQLCommand>
    <SetOptions ANSI_NULLS="ON" ANSI_NULL_DEFAULT="ON" ANSI_PADDING="ON" QUOTED_IDENTIFIER="ON" ENCRYPTED="FALSE" />
    <CommandText>ALTER TABLE dbo.EventTest ADD NewColumn1 INT, NewColumn1 INT;</CommandText>
  </TSQLCommand>
</EVENT_INSTANCE>

因此,由于您可以添加多列,因此您需要检查多列,以获取您可以使用的列列表:

SELECT  ColumnName = Cols.value('.', 'SYSNAME')
FROM    EVENTDATA().nodes('EVENT_INSTANCE/AlterTableActionList/Create/Columns') x (Cols);

然后您可以简单地使用 EXISTS 来检查保留名称列表等,并检查不需要的字符:

IF EXISTS
(   SELECT  1
    FROM    (   SELECT  ColumnName = Cols.value('.', 'SYSNAME')
                FROM    EVENTDATA().nodes('EVENT_INSTANCE/AlterTableActionList/Create/Columns') x (Cols)
            ) AS t
    WHERE   PATINDEX('%[.@!-]%', t.ColumnName) > 0  -- USE PATTERN MATCH TO CHECK FOR UNWANTED CHARACTERS
    OR      EXISTS 
            (   SELECT  1
                FROM    dbo.ReservedColumnNames AS rc
                WHERE   rc.Name = t.ColumnName
            )
)
BEGIN
    RAISERROR(...)
    ROLLBACK;
END

另一件需要考虑的事情是 sp_rename 会触发一个不同的事件,您可能也想跟踪它。这里的 XML 看起来像:

<EVENT_INSTANCE>
  <EventType>RENAME</EventType>
    ....
  <SchemaName>dbo</SchemaName>
  <ObjectName>NewColumn</ObjectName>
  <ObjectType>COLUMN</ObjectType>
  <TargetObjectName>EventTest</TargetObjectName>
  <TargetObjectType>TABLE</TargetObjectType>
  <NewObjectName>NewColumn2</NewObjectName>
  <Parameters>
    <Param>dbo.EventTest.NewColumn</Param>
    <Param>NewColumn2</Param>
    <Param>COLUMN</Param>
  </Parameters>
  <TSQLCommand>
    <SetOptions ANSI_NULLS="ON" ANSI_NULL_DEFAULT="ON" ANSI_PADDING="ON" QUOTED_IDENTIFIER="ON" ENCRYPTED="FALSE" />
    <CommandText>EXECUTE SP_RENAME 'dbo.EventTest.NewColumn', 'NewColumn2', 'COLUMN';</CommandText>
  </TSQLCommand>
</EVENT_INSTANCE>

所以在这里你需要稍微不同地跟踪它,因为它一次只能是一列,你不需要担心节点:

SELECT  EVENTDATA().value('EVENT_INSTANCE[1]/NewObjectName[1]', 'SYSNAME')
WHERE   EVENTDATA().value('EVENT_INSTANCE[1]/EventType[1]', 'VARCHAR(13)') = 'RENAME'
AND     EVENTDATA().value('EVENT_INSTANCE[1]/ObjectType[1]', 'VARCHAR(13)') = 'COLUMN';