正文

SQL Server 2008 存储过程示例(sql server存储过程在哪)速看

文章摘要

以下是关于SQL Server存储过程与函数的简要总结,涵盖了文章中的核心内容: ### SQL Server存储过程与函数总结 1. **存储过程基本语法** - **带输入参数的存储过程**:使用`@`符号接收参数,用于接收外部数据,如`@commentid int`。 ```sql create proc GetComment (@commentid int) as select * from Comment where CommentID=@commentid ``` - **输入与输出参数的存储过程**:使用`@`接收输入参数,并通过`@`输出结果。 ```sql create proc GetCommentCount (@newsid int, @count int output) as select @count=count(*) from Comment where NewsID=@newsid ``` - **返回单个值的函数**:接收参数后返回计算结果。 ```sql create function MyFunction (@newsid int) returns int as begin declare @count int select @count=count(*) from Comment where NewsID=@newsid return @count end ``` - **返回值为表的函数**:直接返回一个结果集。 ```sql create function GetFunctionTable (@newsid int) returns table as return ( select * from Comment where NewsID=@newsid ) ``` 2. **不拼接SQL字符串实现多条件查询** - **传统拼接方法**:易出错,且不推荐。 ```sql set @sql=' select * from table where 1=1 ' if (@addDate is not null) set @sql=@sql+' and addDate='+ @addDate + ' ' ... exec(@sql) ``` - **解决方案**: - 直接写入条件: ```sql select * from table where addDate=... and name=... ``` - 使用`CASE`语句合并条件: ```sql select * from table where addDate=CASE @addDate IS NULL THEN addDate ELSE @addDate END, name=CASE @name WHEN '' THEN name ELSE @name END ``` 3. **SQL Server存储过程与函数应用** - **变量**:包括简单赋值、使用SELECT语句赋值和更新语句赋值。 ```sql declare @a int set @a=5 print @a declare @user1 nvarchar(50) select @user1='张三' print @user1 declare @user2 nvarchar(50) select @user2=Name from ST_User where ID=1 print @user2 ``` - **临时表、表变量与游标**:用于处理中间数据和高效查询。 ```sql -- 创建临时表 create table #DU_User1 ( [ID] [ int ] NOT NULL , [Oid] [ int ] NOT NULL , [Login] [nvarchar](50) NOT NULL , [Rtx] [nvarchar](4) NOT NULL , [ Name ] [nvarchar](5) NOT NULL , [ Password ] [nvarchar]( max ) NULL , [State] [nvarchar](8) NOT NULL ); -- 插入并查询 insert into #DU_User2 from ST_User where ID<8 select * from #DU_User2 where ID<3 union select * from #DU_User1 drop table #DU_User1 drop table #DU_User2 -- �游标实现 declare @ID int declare @Oid int declare @Login varchar(50) declare user_cur cursor for select ID,Oid,[Login] from ST_User open user_cur while @@fetch_status=0 begin fetch next from user_cur into @ID,@Oid,@Login print @ID end close user_cur deallocate user_cur -- 打造游标 declare @ID INT declare @Oid INT declare @Login NVARCHAR(50) declare user_cur cursor for select ID,Oid,[Login] from ST_User open user_cur while @@fetch_status=0 begin fetch next from user_cur into @ID,@Oid,@Login print @ID end close user_cur deallocate user_cur ``` 4. **触发器** - **定义与使用**:用于执行操作前的检查和记录。 ```sql create trigger User_OnUpdate on ST_User for Update as declare @msg nvarchar(50) select @msg=N '姓名从“' + Deleted. Name + '”修改为“' + Inserted. Name + '” from Inserted,Deleted insert into [LOG](MSG) values (@msg) drop trigger User_OnUpdate ``` 5. **存储过程与函数应用示例** - **创建带输出参数的存储过程**: ```sql create proc PR_Sum @a int, @b int, @sum int output as begin set @sum=@a+@b end ``` - **调用存储过程**: ```sql declare @mysum int execute PR_Sum 1,2,@mysum output print @mysum ``` - **自定义函数示例**: ```sql create function FUNC_UserTab_1 ( @myId int ) returns table as return ( select * from ST_User where ID<@myId ) ``` 6. **注意事项** - 使用`CASE`语句合并条件,减少拼接SQL字符串。 - 定义变量和函数,封装业务逻辑,提高复用性和可维护性。 - 使用游标和临时表优化查询性能,避免一次性查询大结果集。 通过以上总结,可以更好地理解和应用SQL Server存储过程和函数,避免传统拼接字符串的缺陷,提升代码质量和效率。


–有输入参数的存储过程– create proc GetComment (@commentid int) as select * from Comment where CommentID=@commentid –有输入与输出参数的存储过程– create proc GetCommentCount @newsid int, @count int output as select @count=count(*) from Comment where NewsID=@newsid –返回单个值的函数– create function MyFunction (@newsid int) returns int as begin declare @count int select @count=count(*) from Comment where NewsID=@newsid return @count end –调用方法– declare @count int exec @count=MyFunction 2 print @count –返回值为表的函数– Create function GetFunctionTable (@newsid int) returns table as return (select * from Comment where NewsID=@newsid) –返回值为表的函数的调用– select * from GetFunctionTable(2)

SQLServer 存储过程中不拼接SQL字符串实现多条件查询

–以前拼接的写法   set @sql=’ select * from table where 1=1 ‘   if (@addDate is not null)    set @sql=@sql+’ and addDate=’+ @addDate + ‘ ‘   if (@name <>” and is not null)    set @sql=@sql+ ‘ and name=’ + @name + ‘ ‘   exec(@sql)

下面是 不采用拼接SQL字符串实现多条件查询的解决方案

  –第一种写法是 感觉代码有些冗余   if (@addDate is not null) and (@name <> ”)    select * from table where addDate=@addDate and name=@name   else if (@addDate is not null) and (@name=”)    select * from table where addDate=@addDate   else if(@addDate is null) and (@name <> ”)    select * from table where and name=@name   else if(@addDate is null) and (@name=”)   select * from table   –第二种写法是   select * from table where (addDate=@addDate or @addDate is null) and (name=@name or @name=”)   –第三种写法是   SELECT * FROM table where   addDate=CASE @addDate IS NULL THEN addDate ELSE @addDate END,   name=CASE @name WHEN ” THEN name ELSE @name END

SQLSERVER存储过程基本语法

一、定义变量

–简单赋值 declare @a int set @a=5 print @a –使用select语句赋值 declare @user1 nvarchar(50) select @user1=’张三’ print @user1 declare @user2 nvarchar(50) select @user2=Name from ST_User where ID=1 print @user2 –使用update语句赋值 declare @user3 nvarchar(50) update ST_User set @user3=Name where ID=1 print @user3

二、表、临时表、表变量

–创建临时表1 create table #DU_User1 ( [ID] [ int ] NOT NULL , [Oid] [ int ] NOT NULL , [Login] [nvarchar](50) NOT NULL , [Rtx] [nvarchar](4) NOT NULL , [ Name ] [nvarchar](5) NOT NULL , [ Password ] [nvarchar]( max ) NULL , [State] [nvarchar](8) NOT NULL ); –向临时表1插入一条记录 insert into #DU_User1 (ID,Oid,[Login],Rtx, Name ,[ Password ],State) values (100,2, ‘LS’ , ‘0000’ , ‘临时’ , ‘321’ , ‘特殊’ ); –从ST_User查询数据,填充至新生成的临时表 select * into #DU_User2 from ST_User where ID<8 –查询并联合两临时表 select * from #DU_User2 where ID<3 union select * from #DU_User1 –删除两临时表 drop table #DU_User1 drop table #DU_User2 –创建临时表 CREATE TABLE #t ( [ID] [ int ] NOT NULL , [Oid] [ int ] NOT NULL , [Login] [nvarchar](50) NOT NULL , [Rtx] [nvarchar](4) NOT NULL , [ Name ] [nvarchar](5) NOT NULL , [ Password ] [nvarchar]( max ) NULL , [State] [nvarchar](8) NOT NULL , ) –将查询结果集(多条数据)插入临时表 insert into #t select * from ST_User –不能这样插入 –select * into #t from dbo.ST_User –添加一列,为int型自增长子段 alter table #t add [myid] int NOT NULL IDENTITY(1,1) –添加一列,默认填充全球唯一标识 alter table #t add [myid1] uniqueidentifier NOT NULL default (newid()) select * from #t drop table #t –给查询结果集增加自增长列 –无主键时: select IDENTITY( int ,1,1) as ID, Name ,[Login],[ Password ] into #t from ST_User select * from #t –有主键时: select ( select SUM (1) from ST_User where ID<=a.ID) as myID,* from ST_User a order by myID –定义表变量 declare @t table ( id int not null , msg nvarchar(50) null ) insert into @t values (1, ‘1’ ) insert into @t values (2, ‘2’ ) select * from @t

三、循环

–while循环计算1到100的和 declare @a int declare @ sum int set @a=1 set @ sum=0 while @a<=100 begin set @ sum +=@a set @a+=1 end print @ sum

四、条件语句

–if,else条件分支 if(1+1=2) begin print ‘对’ end else begin print ‘错’ end –when then条件分支 declare @today int declare @week nvarchar(3) set @today=3 set @week=case when @today=1 then ‘星期一’ when @today=2 then ‘星期二’ when @today=3 then ‘星期三’ when @today=4 then ‘星期四’ when @today=5 then ‘星期五’ when @today=6 then ‘星期六’ when @today=7 then ‘星期日’ else ‘值错误’ end print @week

五、游标

declare @ID int declare @Oid int declare @Login varchar (50) –定义一个游标 declare user_cur cursor for select ID,Oid,[Login] from ST_User –打开游标 open user_cur while @@fetch_status=0 begin –读取游标 fetch next from user_cur into @ID,@Oid,@Login print @ID –print @Login end close user_cur –摧毁游标 deallocate user_cur

五、游标

declare @ID int declare @Oid int declare @Login varchar (50) –定义一个游标 declare user_cur cursor for select ID,Oid,[Login] from ST_User –打开游标 open user_cur while @@fetch_status=0 begin –读取游标 fetch next from user_cur into @ID,@Oid,@Login print @ID –print @Login end close user_cur –摧毁游标 deallocate user_cur

六、触发器

  触发器中的临时表:

  Inserted

  存放进行insert和update 操作后的数据

  Deleted

  存放进行delete 和update操作前的数据
–创建触发器 Create trigger User_OnUpdate On ST_User for Update As declare @msg nvarchar(50) –@msg记录修改情况 select @msg=N ‘姓名从“’ + Deleted. Name + N ‘”修改为“’ + Inserted. Name + ‘”’ from Inserted,Deleted –插入日志表 insert into [LOG](MSG) values (@msg) –删除触发器 drop trigger User_OnUpdate

七、存储过程

–创建带output参数的存储过程 CREATE PROCEDURE PR_Sum @a int , @b int , @ sum int output AS BEGIN set @ sum=@a+@b END –创建Return返回值存储过程 CREATE PROCEDURE PR_Sum2 @a int , @b int AS BEGIN Return @a+@b END –执行存储过程获取output型返回值 declare @mysum int execute PR_Sum 1,2,@mysum output print @mysum –执行存储过程获取Return型返回值 declare @mysum2 int execute @mysum2=PR_Sum2 1,2 print @mysum2

八、自定义函数

  函数的分类:

    1)标量值函数

    2)表值函数

        a:内联表值函数

        b:多语句表值函数

    3)系统函数
–新建标量值函数 create function FUNC_Sum1 ( @a int , @b int ) returns int as begin return @a+@b end –新建内联表值函数 create function FUNC_UserTab_1 ( @myId int ) returns table as return ( select * from ST_User where ID<@myId) –新建多语句表值函数 create function FUNC_UserTab_2 ( @myId int ) returns @t table ( [ID] [ int ] NOT NULL , [Oid] [ int ] NOT NULL , [Login] [nvarchar](50) NOT NULL , [Rtx] [nvarchar](4) NOT NULL , [ Name ] [nvarchar](5) NOT NULL , [ Password ] [nvarchar]( max ) NULL , [State] [nvarchar](8) NOT NULL ) as begin insert into @t select * from ST_User where ID<@myId return end –调用表值函数 select * from dbo.FUNC_UserTab_1(15) –调用标量值函数 declare @s int set @s=dbo.FUNC_Sum1(100,50) print @s –删除标量值函数 drop function FUNC_Sum1
您可能感兴趣的文章:SqlServer获取存储过程返回值的实例SQL SERVER调用存储过程小结一些SQL Server存储过程参数及例子SQLServer用存储过程实现插入更新数据示例SQL Server 分页查询存储过程代码sqlserver存储过程语法详解SQLserver存储过程写法与设置定时执行存储过程方法详解