我需要编写一个T-SQL存储过程来更新表中的一行。如果该行不存在，则插入它。所有这些步骤都由一个事务包装。
这是一个预订系统，所以它必须是原子和可靠的。如果事务已提交且机票已预订，则必须返回true。
我确定如何使用@@rowcount。这就是我到目前为止所写的。我走的路对吗?
-- BEGIN TRANSACTION (HOW TO DO?)
UPDATE Bookings
SET TicketsBooked = TicketsBooked + @TicketsToBook
WHERE FlightId = @Id AND TicketsMax < (TicketsBooked + @TicketsToBook)
-- Here I need to insert only if the row doesn't exists.
-- If the row exists but the condition TicketsMax is violated, I must not insert
-- the row and return FALSE
IF @@ROWCOUNT = 0
BEGIN
INSERT INTO Bookings ... (omitted)
END
-- END TRANSACTION (HOW TO DO?)
-- Return TRUE (How to do?)
看一下MERGE命令。你可以在一条语句中执行UPDATE, INSERT和DELETE操作。
下面是一个使用MERGE的工作实现
-在更新之前检查航班是否已满，否则执行插入操作。
if exists(select 1 from INFORMATION_SCHEMA.TABLES T
where T.TABLE_NAME = 'Bookings')
begin
drop table Bookings
end
GO
create table Bookings(
FlightID int identity(1, 1) primary key,
TicketsMax int not null,
TicketsBooked int not null
)
GO
insert Bookings(TicketsMax, TicketsBooked) select 1, 0
insert Bookings(TicketsMax, TicketsBooked) select 2, 2
insert Bookings(TicketsMax, TicketsBooked) select 3, 1
GO
select * from Bookings
然后……
declare @FlightID int = 1
declare @TicketsToBook int = 2
--; This should add a new record
merge Bookings as T
using (select @FlightID as FlightID, @TicketsToBook as TicketsToBook) as S
on T.FlightID = S.FlightID
and T.TicketsMax > (T.TicketsBooked + S.TicketsToBook)
when matched then
update set T.TicketsBooked = T.TicketsBooked + S.TicketsToBook
when not matched then
insert (TicketsMax, TicketsBooked)
values(S.TicketsToBook, S.TicketsToBook);
select * from Bookings
看一下MERGE命令。你可以在一条语句中执行UPDATE, INSERT和DELETE操作。
下面是一个使用MERGE的工作实现
-在更新之前检查航班是否已满，否则执行插入操作。
if exists(select 1 from INFORMATION_SCHEMA.TABLES T
where T.TABLE_NAME = 'Bookings')
begin
drop table Bookings
end
GO
create table Bookings(
FlightID int identity(1, 1) primary key,
TicketsMax int not null,
TicketsBooked int not null
)
GO
insert Bookings(TicketsMax, TicketsBooked) select 1, 0
insert Bookings(TicketsMax, TicketsBooked) select 2, 2
insert Bookings(TicketsMax, TicketsBooked) select 3, 1
GO
select * from Bookings
然后……
declare @FlightID int = 1
declare @TicketsToBook int = 2
--; This should add a new record
merge Bookings as T
using (select @FlightID as FlightID, @TicketsToBook as TicketsToBook) as S
on T.FlightID = S.FlightID
and T.TicketsMax > (T.TicketsBooked + S.TicketsToBook)
when matched then
update set T.TicketsBooked = T.TicketsBooked + S.TicketsToBook
when not matched then
insert (TicketsMax, TicketsBooked)
values(S.TicketsToBook, S.TicketsToBook);
select * from Bookings
在测试行是否存在时，传递updlock、rowlock、holdlock提示。
begin tran /* default read committed isolation level is fine */
if not exists (select * from Table with (updlock, rowlock, holdlock) where ...)
/* insert */
else
/* update */
commit /* locks are released here */
updlock提示强制查询对已经存在的行使用更新锁，从而防止其他事务在提交或回滚之前修改它。
holdlock提示强制查询接受范围锁，防止其他事务添加匹配您的筛选条件的行，直到您提交或回滚。
rowlock提示将锁粒度强制到行级别，而不是默认的页级别，因此您的事务不会阻止试图更新同一页中不相关行的其他事务(但要注意减少争用和增加锁开销之间的权衡—您应该避免在单个事务中使用大量的行级别锁)。
更多信息请参见http://msdn.microsoft.com/en-us/library/ms187373.aspx。
注意，锁是在执行执行锁的语句时被接受的——调用begin tran并不能让你免受另一个事务的影响，在你到达某个事物之前就对它进行锁定。您应该通过尽可能快地提交事务(晚获取，早释放)来尝试将SQL分解为尽可能短的锁持有时间。
请注意，如果PK是bigint，行级锁可能不太有效，因为SQL Server上的内部哈希对于64位值是退化的(不同的键值可能哈希到相同的锁id)。
这是我最近不得不做的事情:
set ANSI_NULLS ON
set QUOTED_IDENTIFIER ON
GO
ALTER PROCEDURE [dbo].[cjso_UpdateCustomerLogin]
(
@CustomerID AS INT,
@UserName AS VARCHAR(25),
@Password AS BINARY(16)
)
AS
BEGIN
IF ISNULL((SELECT CustomerID FROM tblOnline_CustomerAccount WHERE CustomerID = @CustomerID), 0) = 0
BEGIN
INSERT INTO [tblOnline_CustomerAccount] (
[CustomerID],
[UserName],
[Password],
[LastLogin]
) VALUES (
/* CustomerID - int */ @CustomerID,
/* UserName - varchar(25) */ @UserName,
/* Password - binary(16) */ @Password,
/* LastLogin - datetime */ NULL )
END
ELSE
BEGIN
UPDATE [tblOnline_CustomerAccount]
SET UserName = @UserName,
Password = @Password
WHERE CustomerID = @CustomerID
END
END
