UPDATE() T-SQL 触发函数中的局部变量

Question

我正在使用触发器来更新 table 的更改日志 table，其中包含销售信息。当有人试图在 table 上进行更新时，他们必须写一个描述，然后更新被记录（用户、时间、table、字段、Table_Unique_ID、旧值、新值）

我正在尝试对局部变量@fieldname 使用 UPDATE() 触发函数。这样我的目标是只比较更新的列的 deleted/inserted 值，而不是循环检查每一列。当我明确声明列名时，我已经能够使用 UPDATE() 函数，但如果我将变量设置为我要检查的列名，则不能。

我非常感谢有关如何优化此过程的任何建议！我还有很多东西要学，这是我第一次尝试记录更改。

下面是我的完整触发器：

ALTER TRIGGER [dbo].[TR_UpdateLog]

 ON [dbo].[Test_Update_Trigger]
    AFTER  Update
 AS

Declare @description nvarchar(1000)
, @UserName nvarchar(128)
, @oldValue nvarchar(255)
, @newValue nvarchar(255)
, @UniqueID nvarchar(255)
, @fieldname nvarchar(128)
, @oname NVARCHAR(100)
, @OldSQL nvarchar(max)
, @NewSQL nvarchar(max)
, @i int
, @c int
, @numcolumns int
, @numrows int;
DECLARE @updated_table TABLE ( 
idx int Primary Key identity(1,1)
, uniqueID nvarchar(255) NULL )
;


--Require the User to submit a description for the update
Set @description = (SELECT * FROM [dbo].[Fact_Bookings_Audit_Description])
TRUNCATE TABLE [dbo].[Fact_Bookings_Audit_Description]

IF @description is null 
BEGIN
    PRINT 'You must provide a description. Use the following text:

    TRUNCATE TABLE [dbo].[Fact_Bookings_Audit_Description]
    INSERT INTO [dbo].[Fact_Bookings_Audit_Description]
    SELECT "Your Text Here--use Single Quotes"

    Copy and paste above your query, type your description, and run the update again.
    ';
    ROLLBACK TRANSACTION;
END

--Set User and Table name
Set @Username = system_User
SET @oname = (SELECT OBJECT_NAME(parent_id) from sys.triggers where object_ID = @@PROCID)


-------------------------------------------------------------
---create tables of updated items to reference for old/new values
INSERT INTO @updated_table
SELECT distinct Table_ID from DELETED

SELECT * INTO dbo.tempDelete FROM DEleted 
SELECT * INTO dbo.tempInsert FROM inserted
--------------------------------------------------
--Set variables for loop through updated items
Set @i  = 1
Set @numrows = (SELECT count(*) from @updated_table)
------------------------------------------------------------
---Set number of columns for loop through each field
Set @numcolumns = (SELECT MAx(ordinal_position) FROM information_schema.columns where table_name = @oname )

---------------------------------------------------------------
--If there was an update
IF @numrows > 0
BEGIN
    --loop through each individual updated row
    WHILE (@i <= (select max(idx) from @updated_table))
    BEGIN
        --reset the column variable
        Set @c = 1
        --loop through each column
        WHILE (@c <= @numcolumns)
        BEGIN
                Set @fieldname =    (SELECT Column_name FROM information_schema.columns 
                                        Where table_name = @oname AND ORDINAL_POSITION = @c)

                    /**This is what I want to use--
                    I would only like to do the comparison on columns which were updated, not cycle through every column**/

                    --IF UPDATE(@fieldname)
                    --BEGIN

                    --set values for old and new values as well as unique ID for log table
                    Set @UniqueID = (SELECT uniqueID from @updated_table u WHERE idx = @i)
                    Set @OldSQL = 'SELECT @oldvalue = ' + @fieldname + ' from dbo.tempdelete d WHERE d.Table_ID = ' + @UniqueID;
                    Set @NewSQL = 'SELECT @newvalue = ' + @fieldname + ' from dbo.tempInsert i WHERE i.Table_ID = ' + @UniqueID;

                    EXEC sp_executesql @oldSQL, N'@oldvalue nvarchar(128) output', @oldValue = @oldValue output
                    EXEC sp_executesql @newSQL, N'@newvalue nvarchar(128) output', @newValue = @newValue output

                    ;
                        --Insert into log table if value is changed
                        IF isnull(@oldvalue,0) <> isnull(@newvalue,0)
                        BEGIN
                            INSERT INTO dbo.Fact_Bookings_ChangeAudit (Change_Date, Change_Time, [User], Table_Name, Field, Table_ID, Oldvalue, Newvalue, [Description])
                            VALUES(cast(datediff(DAY,0,getdate())as datetime),cast(getdate() as datetime),@UserName, @oname, @fieldname, @UniqueID, @oldValue, @newValue, @description);
                        END
                    --END
                --next column
                Set @c = @c + 1
        END
    --next record
    Set @i = @i + 1
    END

END
DROP TABLE dbo.tempDelete
DROP table dbo.tempInsert

GO

Answer 1

with i as (
    select
        case when updated(colA)
            then coalesce(cast(ColA as varchar(32)), 'null') end as newColA,
        case when updated(colB)
            then coalesce(cast(ColB as varchar(32)), 'null') end as newColB,
        ...
    from inserted
), d as (
    select
        case when updated(colA)
            then coalesce(cast(ColA as varchar(32)), 'null') end as oldColA,
        case when updated(colB)
            then coalesce(cast(ColB as varchar(32)), 'null') end as oldColB,
        ...
    from deleted
), old as (
    select c.Id, Col, Change
    from i unpivot (Change for Col in (newColA, newColB, ...))
    where Change is not null
), new as (
    select c.Id, Col, Change
    from d unpivot (Change for Col in (oldColA, oldColB, ...))
    where Change is not null
)
-- insert into Log
select ...
from
    old o inner join new n on n.Id = o.Id and n.Col = o.Col

列出列是执行此操作的正确方法。我知道您希望对信息模式表使用循环。我只是不认为动态 sql 是个好主意。

我认为上面的查询可能是创建要记录的行列表的起点。我没有尝试测试它。如果您尝试进行批量更新，它可能会很慢。

批量更新可能比使用连接更快？

with ...
, new as (select 'i' as src ...),
, old as (select 'd' as src ...),
, combined as (select * from new union all select * from old)
select
    Id, Col
    min(case when src = 'i' then Change end) as newVal,
    min(case when src = 'd' then Change end) as oldVal,
from combined
group by Id, Col;

UPDATE() T-SQL 触发函数中的局部变量

Local Variable in UPDATE() T-SQL Trigger Function

tsql

sql-server

database-trigger