插入更新触发器如何确定是插入还是更新

我需要在表A上编写一个插入，更新触发器，该操作将从表B的所有行中删除所有行，该表的某一列（例如Desc）的值类似于在表A的列中插入/更新的值（例如Col1）。我将如何编写它，以便同时处理Update和Insert案例。我将如何确定是否为更新或插入执行了触发器。

触发器具有特殊的INSERTED和DELETED表来跟踪“之前”和“之后”数据。因此，您可以使用类似的方法IF EXISTS (SELECT * FROM DELETED)来检测更新。DELETED更新时您只有行，但中总是有行INSERTED。

在CREATE TRIGGER中查找“插入” 。

编辑，2011年11月23日

评论后，此答案仅适用于INSERTED和UPDATED触发。
显然，INSERTED正如我在上文所述，DELETE触发器不能“始终有行”

CREATE TRIGGER dbo.TableName_IUD
ON dbo.TableName
AFTER INSERT, UPDATE, DELETE
AS 
BEGIN
    SET NOCOUNT ON;

    --
    -- Check if this is an INSERT, UPDATE or DELETE Action.
    -- 
    DECLARE @action as char(1);

    SET @action = 'I'; -- Set Action to Insert by default.
    IF EXISTS(SELECT * FROM DELETED)
    BEGIN
        SET @action = 
            CASE
                WHEN EXISTS(SELECT * FROM INSERTED) THEN 'U' -- Set Action to Updated.
                ELSE 'D' -- Set Action to Deleted.       
            END
    END
    ELSE 
        IF NOT EXISTS(SELECT * FROM INSERTED) RETURN; -- Nothing updated or inserted.

    ...

    END

如果您运行不删除任何内容的delete语句，则许多建议都不会考虑在内。
假设您尝试删除ID等于表中不存在的某个值的地方。
您的触发器仍然被调用，但是“已删除”或“插入”表中没有任何内容。

为了安全起见，请使用以下命令：

--Determine if this is an INSERT,UPDATE, or DELETE Action or a "failed delete".
DECLARE @Action as char(1);
    SET @Action = (CASE WHEN EXISTS(SELECT * FROM INSERTED)
                         AND EXISTS(SELECT * FROM DELETED)
                        THEN 'U'  -- Set Action to Updated.
                        WHEN EXISTS(SELECT * FROM INSERTED)
                        THEN 'I'  -- Set Action to Insert.
                        WHEN EXISTS(SELECT * FROM DELETED)
                        THEN 'D'  -- Set Action to Deleted.
                        ELSE NULL -- Skip. It may have been a "failed delete".   
                    END)

特别感谢@KenDog和@Net_Prog的回答。
我是根据他们的脚本构建的。

我正在使用以下内容，它还可以正确检测不删除任何内容的删除语句：

CREATE TRIGGER dbo.TR_TableName_TriggerName
    ON dbo.TableName
    AFTER INSERT, UPDATE, DELETE
AS
BEGIN
    SET NOCOUNT ON;

    IF NOT EXISTS(SELECT * FROM INSERTED)
        -- DELETE
        PRINT 'DELETE';
    ELSE
    BEGIN
        IF NOT EXISTS(SELECT * FROM DELETED)
            -- INSERT
            PRINT 'INSERT';
        ELSE
            -- UPDATE
            PRINT 'UPDATE';
    END
END;

经过大量搜索，我找不到单个SQL Server触发器的确切示例，该触发器可处理触发器动作INSERT，UPDATE和DELETE的所有（3）三个条件。我终于找到一行文字，讨论当发生DELETE或UPDATE时，公共DELETED表将包含这两个动作的记录这一事实。根据这些信息，我然后创建了一个小的Action例程，该例程确定了为什么激活了触发器。当在INSERT与UPDATE触发器上同时进行通用配置和特定操作时，有时需要这种类型的接口。在这些情况下，为UPDATE和INSERT创建单独的触发器将成为维护问题。（即，是否已针对必要的通用数据算法修复程序正确更新了两个触发器？）

为此，我想提供以下多触发事件代码段，用于在Microsoft SQL Server的一个触发器中处理INSERT，UPDATE，DELETE。

CREATE TRIGGER [dbo].[INSUPDDEL_MyDataTable]
ON [dbo].[MyDataTable] FOR INSERT, UPDATE, DELETE
AS 

-- SET NOCOUNT ON added to prevent extra result sets from
-- interfering with caller queries SELECT statements.
-- If an update/insert/delete occurs on the main table, the number of records affected
-- should only be based on that table and not what records the triggers may/may not
-- select.
SET NOCOUNT ON;

--
-- Variables Needed for this Trigger
-- 
DECLARE @PACKLIST_ID varchar(15)
DECLARE @LINE_NO smallint
DECLARE @SHIPPED_QTY decimal(14,4)
DECLARE @CUST_ORDER_ID varchar(15)
--
-- Determine if this is an INSERT,UPDATE, or DELETE Action
-- 
DECLARE @Action as char(1)
DECLARE @Count as int
SET @Action = 'I' -- Set Action to 'I'nsert by default.
SELECT @Count = COUNT(*) FROM DELETED
if @Count > 0
    BEGIN
        SET @Action = 'D' -- Set Action to 'D'eleted.
        SELECT @Count = COUNT(*) FROM INSERTED
        IF @Count > 0
            SET @Action = 'U' -- Set Action to 'U'pdated.
    END

if @Action = 'D'
    -- This is a DELETE Record Action
    --
    BEGIN
        SELECT @PACKLIST_ID =[PACKLIST_ID]
                    ,@LINE_NO = [LINE_NO]
        FROM DELETED

        DELETE [dbo].[MyDataTable]
        WHERE [PACKLIST_ID]=@PACKLIST_ID AND [LINE_NO]=@LINE_NO
    END
 Else
    BEGIN
            --
            -- Table INSERTED is common to both the INSERT, UPDATE trigger
            --
            SELECT @PACKLIST_ID =[PACKLIST_ID]
                ,@LINE_NO = [LINE_NO]
                ,@SHIPPED_QTY =[SHIPPED_QTY]
                ,@CUST_ORDER_ID = [CUST_ORDER_ID]
            FROM INSERTED 

         if @Action = 'I'
            -- This is an Insert Record Action
            --
            BEGIN
                INSERT INTO [MyChildTable]
                    (([PACKLIST_ID]
                    ,[LINE_NO]
                    ,[STATUS]
                VALUES
                    (@PACKLIST_ID
                    ,@LINE_NO
                    ,'New Record'
                    )
            END
        else
            -- This is an Update Record Action
            --
            BEGIN
                UPDATE [MyChildTable]
                    SET [PACKLIST_ID] = @PACKLIST_ID
                          ,[LINE_NO] = @LINE_NO
                          ,[STATUS]='Update Record'
                WHERE [PACKLIST_ID]=@PACKLIST_ID AND [LINE_NO]=@LINE_NO
            END
    END   

我相信嵌套ifs会造成一些混乱，并且：

扁平比嵌套更好[Python的禅宗]

;）

DROP TRIGGER IF EXISTS AFTER_MYTABLE

GO

CREATE TRIGGER dbo.AFTER_MYTABLE ON dbo.MYTABLE AFTER INSERT, UPDATE, DELETE 

AS BEGIN 

    --- FILL THE BEGIN/END SECTION FOR YOUR NEEDS.

    SET NOCOUNT ON;

    IF EXISTS(SELECT * FROM INSERTED)  AND EXISTS(SELECT * FROM DELETED) 
        BEGIN PRINT 'UPDATE' END 
    ELSE IF EXISTS(SELECT * FROM INSERTED)  AND NOT EXISTS(SELECT * FROM DELETED) 
        BEGIN PRINT 'INSERT' END 
    ELSE IF    EXISTS(SELECT * FROM DELETED) AND NOT EXISTS(SELECT * FROM INSERTED)
        BEGIN PRINT 'DELETED' END
    ELSE BEGIN PRINT 'NOTHING CHANGED'; RETURN; END  -- NOTHING

END

Declare @Type varchar(50)='';
IF EXISTS (SELECT * FROM inserted) and  EXISTS (SELECT * FROM deleted)
BEGIN
    SELECT @Type = 'UPDATE'
END
ELSE IF EXISTS(SELECT * FROM inserted)
BEGIN
    SELECT @Type = 'INSERT'
END
ElSE IF EXISTS(SELECT * FROM deleted)
BEGIN
    SELECT @Type = 'DELETE'
END

试试这个..

ALTER TRIGGER ImportacionesGS ON dbo.Compra 
    AFTER INSERT, UPDATE, DELETE
AS
BEGIN
  -- idCompra is PK
  DECLARE @vIdCompra_Ins INT,@vIdCompra_Del INT
  SELECT @vIdCompra_Ins=Inserted.idCompra FROM Inserted
  SELECT @vIdCompra_Del=Deleted.idCompra FROM Deleted
  IF (@vIdCompra_Ins IS NOT NULL AND @vIdCompra_Del IS NULL)  
  Begin
     -- Todo Insert
  End
  IF (@vIdCompra_Ins IS NOT NULL AND @vIdCompra_Del IS NOT NULL)
  Begin
     -- Todo Update
  End
  IF (@vIdCompra_Ins IS NULL AND @vIdCompra_Del IS NOT NULL)
  Begin
     -- Todo Delete
  End
END

虽然我也喜欢@Alex发布的答案，但我在上面提供了@Graham解决方案的这种变化

与在第一个测试中使用COLUMNS_UPDATED相反，这专门使用了INSERTED和UPDATED表中的记录存在。知道已经考虑了最后一种情况，这也使程序员感到疑惑。

declare @action varchar(4)
    IF EXISTS (SELECT * FROM INSERTED)
        BEGIN
            IF EXISTS (SELECT * FROM DELETED) 
                SET @action = 'U'  -- update
            ELSE
                SET @action = 'I'  --insert
        END
    ELSE IF EXISTS (SELECT * FROM DELETED)
        SET @action = 'D'  -- delete
    else 
        set @action = 'noop' --no records affected
--print @action

您将通过以下语句获得NOOP：

update tbl1 set col1='cat' where 1=2

这可能是一种更快的方法：

DECLARE @action char(1)

IF COLUMNS_UPDATED() > 0 -- insert or update
BEGIN
    IF EXISTS (SELECT * FROM DELETED) -- update
        SET @action = 'U'
    ELSE
        SET @action = 'I'
    END
ELSE -- delete
    SET @action = 'D'

所提供的两种解决方案的潜在问题是，取决于如何编写，更新查询可能会更新零个记录，而插入查询可能会插入零个记录。在这种情况下，插入和删除记录集将为空。在许多情况下，如果“插入”和“已删除”记录集均为空，则可能只想退出触发器而无需执行任何操作。

我在Grahams的其他不错的解决方案中发现了一个小错误：

应该为IF COLUMNS_UPDATED（）< > 0-插入或更新
而不是> 0，这可能是因为最高位被解释为SIGNED整数符号位...（？）。因此总计：

DECLARE @action CHAR(8)  
IF COLUMNS_UPDATED() <> 0 -- delete or update?
BEGIN     
  IF EXISTS (SELECT * FROM deleted) -- updated cols + old rows means action=update       
    SET @action = 'UPDATE'     
  ELSE
    SET @action = 'INSERT' -- updated columns and nothing deleted means action=insert
END 
ELSE -- delete     
BEGIN
  SET @action = 'DELETE'
END

这对我有用：

declare @action_type int;
select @action_type = case
                       when i.id is not null and d.id is     null then 1 -- insert
                       when i.id is not null and d.id is not null then 2 -- update
                       when i.id is     null and d.id is not null then 3 -- delete
                     end
  from      inserted i
  full join deleted  d on d.id = i.id

由于并非一次可以更新所有列，因此您可以检查某个特定列是否正在通过以下方式进行更新：

IF UPDATE([column_name])

declare @insCount int
declare @delCount int
declare @action char(1)

select @insCount = count(*) from INSERTED
select @delCount = count(*) from DELETED

    if(@insCount > 0 or @delCount > 0)--if something was actually affected, otherwise do nothing
    Begin
        if(@insCount = @delCount)
            set @action = 'U'--is update
        else if(@insCount > 0)
            set @action = 'I' --is insert
        else
            set @action = 'D' --is delete

        --do stuff here
    End

我喜欢“计算机科学优雅”的解决方案。我的解决方案在这里分别命中[插入]和[删除]伪表以获取其状态，并将结果放入位映射变量中。然后，可以通过有效的二进制评估轻松地在整个触发器中测试INSERT，UPDATE和DELETE的每种可能组合（除非不太可能的INSERT或DELETE组合）。

它确实假设，如果不修改任何行，DML语句是什么无关紧要（这应满足绝大多数情况）。因此，尽管它不如Roman Pekar的解决方案完整，但效率更高。

通过这种方法，我们可以在每个表中使用一个“ FOR INSERT，UPDATE，DELETE”触发器，这使我们A）可以完全控制操作顺序，b）可以执行每个适用于多操作的代码。（显然，每个实现模型都有其优点和缺点；您将需要分别评估系统以找出最有效的方法。）

请注意，“存在（从插入/删除中选择*）”语句非常有效，因为没有磁盘访问权限（https://social.msdn.microsoft.com/Forums/en-US/01744422-23fe-42f6 -9ab0-a255cdf2904a）。

use tempdb
;
create table dbo.TrigAction (asdf int)
;
GO
create trigger dbo.TrigActionTrig
on dbo.TrigAction
for INSERT, UPDATE, DELETE
as
declare @Action tinyint
;
-- Create bit map in @Action using bitwise OR "|"
set @Action = (-- 1: INSERT, 2: DELETE, 3: UPDATE, 0: No Rows Modified 
  (select case when exists (select * from inserted) then 1 else 0 end)
| (select case when exists (select * from deleted ) then 2 else 0 end))
;
-- 21 <- Binary bit values
-- 00 -> No Rows Modified
-- 01 -> INSERT -- INSERT and UPDATE have the 1 bit set
-- 11 -> UPDATE <
-- 10 -> DELETE -- DELETE and UPDATE have the 2 bit set

raiserror(N'@Action = %d', 10, 1, @Action) with nowait
;
if (@Action = 0) raiserror(N'No Data Modified.', 10, 1) with nowait
;
-- do things for INSERT only
if (@Action = 1) raiserror(N'Only for INSERT.', 10, 1) with nowait
;
-- do things for UPDATE only
if (@Action = 3) raiserror(N'Only for UPDATE.', 10, 1) with nowait
;
-- do things for DELETE only
if (@Action = 2) raiserror(N'Only for DELETE.', 10, 1) with nowait
;
-- do things for INSERT or UPDATE
if (@Action & 1 = 1) raiserror(N'For INSERT or UPDATE.', 10, 1) with nowait
;
-- do things for UPDATE or DELETE
if (@Action & 2 = 2) raiserror(N'For UPDATE or DELETE.', 10, 1) with nowait
;
-- do things for INSERT or DELETE (unlikely)
if (@Action in (1,2)) raiserror(N'For INSERT or DELETE.', 10, 1) with nowait
-- if already "return" on @Action = 0, then use @Action < 3 for INSERT or DELETE
;
GO

set nocount on;

raiserror(N'
INSERT 0...', 10, 1) with nowait;
insert dbo.TrigAction (asdf) select top 0 object_id from sys.objects;

raiserror(N'
INSERT 3...', 10, 1) with nowait;
insert dbo.TrigAction (asdf) select top 3 object_id from sys.objects;

raiserror(N'
UPDATE 0...', 10, 1) with nowait;
update t set asdf = asdf /1 from dbo.TrigAction t where asdf <> asdf;

raiserror(N'
UPDATE 3...', 10, 1) with nowait;
update t set asdf = asdf /1 from dbo.TrigAction t;

raiserror(N'
DELETE 0...', 10, 1) with nowait;
delete t from dbo.TrigAction t where asdf < 0;

raiserror(N'
DELETE 3...', 10, 1) with nowait;
delete t from dbo.TrigAction t;
GO

drop table dbo.TrigAction
;
GO

快速解决方案MySQL

顺便说一句：我正在使用MySQL PDO。

（1）在自动增量表中，每个脚本首先运行一次，只需从增量列中获取最大值（我的列名称= id）：

$select = "
    SELECT  MAX(id) AS maxid
    FROM    [tablename]
    LIMIT   1
";

（2）像往常一样运行MySQL查询，并将结果转换为整数，例如：

$iMaxId = (int) $result[0]->maxid;

（3）在“ INSERT INTO ... ON DUPLICATE KEY UPDATE”查询之后，以您喜欢的方式获取最后插入的ID，例如：

$iLastInsertId = (int) $db->lastInsertId();

（4）比较并作出反应：如果lastInsertId高于表中的最高值，则可能是INSERT，对吗？反之亦然。

if ($iLastInsertId > $iMaxObjektId) {
    // IT'S AN INSERT
}
else {
    // IT'S AN UPDATE
}

我知道它很快，甚至很脏。这是一个旧帖子。但是，嘿，我长期以来一直在寻找解决方案，也许有人发现我的方式仍然很有用。祝一切顺利！

只是简单的方法

CREATE TRIGGER [dbo].[WO_EXECUTION_TRIU_RECORD] ON [dbo].[WO_EXECUTION]
WITH EXECUTE AS CALLER
FOR INSERT, UPDATE
AS
BEGIN  

  select @vars = [column] from inserted 
  IF UPDATE([column]) BEGIN
    -- do update action base on @vars 
  END ELSE BEGIN
    -- do insert action base on @vars 
  END

END 

在第一种情况下，我认为您的表具有IDENTITY列

CREATE TRIGGER [dbo].[insupddel_yourTable] ON [yourTable]
FOR INSERT, UPDATE, DELETE
AS
IF @@ROWCOUNT = 0 return
SET NOCOUNT ON;
DECLARE @action nvarchar(10)
SELECT @action = CASE WHEN COUNT(i.Id) > COUNT(d.Id) THEN 'inserted'
                      WHEN COUNT(i.Id) < COUNT(d.Id) THEN 'deleted' ELSE 'updated' END
FROM inserted i FULL JOIN deleted d ON i.Id = d.Id

在第二种情况下，不需要使用IDENTITTY列

CREATE TRIGGER [dbo].[insupddel_yourTable] ON [yourTable]
FOR INSERT, UPDATE, DELETE
AS
IF @@ROWCOUNT = 0 return
SET NOCOUNT ON;
DECLARE @action nvarchar(10),
        @insCount int = (SELECT COUNT(*) FROM inserted),
        @delCount int = (SELECT COUNT(*) FROM deleted)
SELECT @action = CASE WHEN @insCount > @delCount THEN 'inserted'
                      WHEN @insCount < @delCount THEN 'deleted' ELSE 'updated' END

DECLARE @INSERTEDCOUNT INT,
        @DELETEDCOUNT INT

SELECT @INSERTEDCOUNT = COUNT([YourColumnName]) FROM inserted

SELECT @DELETEDCOUNT = COUNT([YourColumnName]) FROM deleted

如果更新

 @INSERTEDCOUNT = 1
 @DELETEDCOUNT = 1

如果插入

 @INSERTEDCOUNT = 1
 @DELETEDCOUNT = 0

我已经使用这些exists (select * from inserted/deleted)查询很长时间了，但是对于空的CRUD操作（当insertedand deleted表中没有记录时）仍然不够。因此，在研究了这一主题后，我找到了更精确的解决方案：

declare
    @columns_count int = ?? -- number of columns in the table,
    @columns_updated_count int = 0

-- this is kind of long way to get number of actually updated columns
-- from columns_updated() mask, it's better to create helper table
-- or at least function in the real system
with cte_columns as (
    select @columns_count as n
    union all
    select n - 1 from cte_columns where n > 1
), cte_bitmasks as (
    select
        n,
        (n - 1) / 8 + 1 as byte_number,
        power(2, (n - 1) % 8) as bit_mask
    from cte_columns
)
select
    @columns_updated_count = count(*)
from cte_bitmasks as c
where
    convert(varbinary(1), substring(@columns_updated_mask, c.byte_number, 1)) & c.bit_mask > 0

-- actual check
if exists (select * from inserted)
    if exists (select * from deleted)
        select @operation = 'U'
    else
        select @operation = 'I'
else if exists (select * from deleted)
    select @operation = 'D'
else if @columns_updated_count = @columns_count
    select @operation = 'I'
else if @columns_updated_count > 0
    select @operation = 'U'
else
    select @operation = 'D'

也可以使用 columns_updated() & power(2, column_id - 1) > 0查看列是否已更新，但是对于具有大量列的表来说并不安全。我使用了一种复杂的计算方式（请参阅下面的帮助文章）。

同样，此方法仍会错误地将某些更新分类为插入（如果表中的每一列都受到更新的影响），并且可能会将仅默认值插入为删除的插入分类，但这些是罕见操作的王者（在租赁在我的系统中）。除此之外，我目前不知道如何改进此解决方案。

• COLUMNS_UPDATED（）用于Piotr Rodak的审计触发器
• 亚当·麦克汉尼（Adam Machanic）处理非常大的位掩码

declare @result as smallint
declare @delete as smallint = 2
declare @insert as smallint = 4
declare @update as smallint = 6
SELECT @result = POWER(2*(SELECT count(*) from deleted),1) + POWER(2*(SELECT 
     count(*) from inserted),2)

if (@result & @update = @update) 
BEGIN
  print 'update'
  SET @result=0
END
if (@result & @delete = @delete)
  print 'delete'
if (@result & @insert = @insert)
  print 'insert'

我这样做：

select isnull((select top 1 1 from inserted t1),0) + isnull((select top 1 2 from deleted t1),0)

1->插入

2->删除

3->更新

set @i = isnull((select top 1 1 from inserted t1),0) + isnull((select top 1 2 from deleted t1),0)
--select @i

declare @action varchar(1) = case @i when 1 then 'I' when 2 then 'D' when 3 then 'U' end
--select @action


select @action c1,* from inserted t1 where @i in (1,3) union all
select @action c1,* from deleted t1 where @i in (2)

DECLARE @ActionType CHAR(6);
SELECT  @ActionType = COALESCE(CASE WHEN EXISTS(SELECT * FROM INSERTED)
                                     AND EXISTS(SELECT * FROM DELETED)  THEN 'UPDATE' END,
                               CASE WHEN EXISTS(SELECT * FROM DELETED)  THEN 'DELETE' END,
                               CASE WHEN EXISTS(SELECT * FROM INSERTED) THEN 'INSERT' END);
PRINT   @ActionType;