本文转自:http://geekswithblogs.net/bbiales/archive/2012/03/15/how-to-nest-transactions-nicely---quotbegin-transactionquot-vs-quotsave.aspx

Do you write stored procedures that might be used by others?  And those others may or may not have already started a transaction?  And your SP does several things, but if any of them fail, you have to undo them all and return with a code indicating it failed?

Well, I have written such code, and it wasn’t working right until I finally figured out how to handle the case when we are already in a transaction, as well as the case where the caller did not start a transaction.  When a problem occurred, my “ROLLBACK TRANSACTION” would roll back not just my nested transaction, but the caller’s transaction as well.  So when I tested the procedure stand-alone, it seemed to work fine, but when others used it, it would cause a problem if it had to rollback.  When something went wrong in my procedure, their entire transaction was rolled back.  This was not appreciated.

Now, I knew one could "nest" transactions, but the technical documentation was very confusing.  And I still have not found the approach below documented anywhere.  So here is a very brief description of how I got it to work, I hope you find this helpful.

My example is a stored procedure that must figure out on its own if the caller has started a transaction or not.  This can be done in SQL Server by checking the @@TRANCOUNT value.  If no BEGIN TRANSACTION has occurred yet, this will have a value of 0.  Any number greater than zero means that a transaction is in progress.  If there is no current transaction, my SP begins a transaction. But if a transaction is already in progress, my SP uses SAVE TRANSACTION and gives it a name.  SAVE TRANSACTION creates a “save point”.  Note that creating a save point has no effect on @@TRANCOUNT.

So my SP starts with something like this:

DECLARE @startingTranCount int
SET @startingTranCount = @@TRANCOUNT IF @startingTranCount > 0
SAVE TRANSACTION mySavePointName
ELSE
BEGIN TRANSACTION
-- …


Then, when ready to commit the changes, you only need to commit if we started the transaction ourselves:

IF @startingTranCount = 0
COMMIT TRANSACTION

And finally, to roll back just your changes so far:

-- Roll back changes...
IF @startingTranCount > 0
ROLLBACK TRANSACTION MySavePointName
ELSE
ROLLBACK TRANSACTION

Here is some code that you can try that will demonstrate how the save points work inside a transaction.

This sample code creates a temporary table, then executes selects and updates, documenting what is going on, then deletes the temporary table.

if running in SQL Management Studio, set Query Results to: Text for best readability of the results.

-- Create a temporary table to test with, we'll drop it at the end.
CREATE TABLE #ATable(
[Column_A] [varchar](5) NULL
) ON [PRIMARY] GO
SET NOCOUNT ON
-- Ensure just one row - delete all rows, add one
DELETE #ATable
-- Insert just one row
INSERT INTO #ATable VALUES('000') SELECT 'Before TRANSACTION starts, value in table is: ' AS Note, * FROM #ATable SELECT @@trancount AS CurrentTrancount
--insert into a values ('abc')
UPDATE #ATable SET Column_A = 'abc'
SELECT 'UPDATED without a TRANSACTION, value in table is: ' AS Note, * FROM #ATable
BEGIN TRANSACTION
SELECT 'BEGIN TRANSACTION, trancount is now ' AS Note, @@TRANCOUNT AS TranCount
UPDATE #ATable SET Column_A = '123'
SELECT 'Row updated inside TRANSACTION, value in table is: ' AS Note, * FROM #ATable
SAVE TRANSACTION MySavepoint
SELECT 'Save point MySavepoint created, transaction count now:' as Note, @@TRANCOUNT AS TranCount
UPDATE #ATable SET Column_A = '456'
SELECT 'Updated after MySavepoint created, value in table is: ' AS Note, * FROM #ATable
SAVE TRANSACTION point2
SELECT 'Save point point2 created, transaction count now:' as Note, @@TRANCOUNT AS TranCount
UPDATE #ATable SET Column_A = '789'
SELECT 'Updated after point2 savepoint created, value in table is: ' AS Note, * FROM #ATable
ROLLBACK TRANSACTION point2
SELECT 'Just rolled back savepoint "point2", value in table is: ' AS Note, * FROM #ATable
ROLLBACK TRANSACTION MySavepoint
SELECT 'Just rolled back savepoint "MySavepoint", value in table is: ' AS Note, * FROM #ATable
SELECT 'Both save points were rolled back, transaction count still:' as Note, @@TRANCOUNT AS TranCount
ROLLBACK TRANSACTION
SELECT 'Just rolled back the entire transaction..., value in table is: ' AS Note, * FROM #ATable DROP TABLE #ATable

[转]How to nest transactions nicely - "begin transaction" vs "save transaction" and SQL Server的更多相关文章

  1. 事务操作(BEGIN/COMMIT/ROLLBACK/SAVE TRANSACTION)

    BEGIN TRANSACTION 标记一个显式本地事务的起始点. BEGIN TRANSACTION 使 @@TRANCOUNT 按 1 递增. BEGIN TRANSACTION 代表一点,由连接 ...

  2. SQL SAVE TRANSACTION

    --创建存储过程 create procedure qiantaoProc @asd nchar(10) as begin begin try begin transaction innerTrans ...

  3. SQL Server does not purge row versioning records even the transaction are committed if there are other open transaction running in the databases with read-committed snapshot enabled .

    This is a by-design behavior. There is only one allocation unit in tempdb that istracking the versio ...

  4. [转]sql server transaction

    本文转自: http://www.2cto.com/database/201208/146734.html sql事务(Transaction)用法介绍及回滚实例   事务(Transaction)是 ...

  5. sql server 学习笔记 (nested transaction 嵌套事务)

    什么时候会用到嵌套事务 ? 为了代码复用,我们会写许多的储蓄过程,而中间如果需要使用到 transaction 难免就会发生嵌套了. sql server 并不直接支持嵌套事务. 但它可以用一些招式来 ...

  6. SQL Server 数据库的维护(三)__事务(transaction)和锁

    --维护数据库-- --事务(transaction)和锁-- --事务(transaction)-- --概述: 事务是指封装了一组T-SQL语句的单个逻辑单元.单元中的所有语句作为一个整体,在满足 ...

  7. [杂]SQL Server 之 Understanding Connection Pooling and Transactions

    A SqlConnection consists of two parts: the public instance that your code interacts with (the outer ...

  8. SQL Server save transaction

    准备: create table Nums(X int); 目的:只向表中插入一行. --------------------------------------------------------- ...

  9. [官网]How to use distributed transactions with SQL Server on Docker

    How to use distributed transactions with SQL Server on Docker https://docs.microsoft.com/en-us/sql/l ...

随机推荐

  1. 设计模式之代理模式(Proxy Pattern)_补充篇

    写在前面: 代理模式的内部原理,作用及远程代理的实现在上一篇博文中都做了详细解释,本文只是对其内容的补充,介绍其它代理 一.虚拟代理 首先,明确虚拟代理的作用:在巨大对象被真正创建出来之前,用虚拟代理 ...

  2. MongoDB Windowns 配置使用

     MongoDB 下载 MongoDB 提供了可用于 32 位和 64 位系统的预编译二进制包,你可以从MongoDB官网下载安装,MongoDB 预编译二进制包下载地址:https://www.mo ...

  3. .net core 生成二维码

    其实生成二维码的组件有很多种,如:QrcodeNet,ZKWeb.Fork.QRCoder,QRCoder等 我选QRCoder,是因为小而易用.支持大并发生成请求.不依赖任何库和网络服务. 既然是. ...

  4. 利用ligerUI隐藏某列,并不产生空白列的方法

    var grid;//声明变量 $(function () { //grid初始化 grid = $("#maingrid4").ligerGrid({ columns: [ { ...

  5. kubenetes--Namespace命名空间

    Namespace(命名空间)是kubernetes系统中的另一个重要的概念,通过将系统内部的对象“分配”到不同的Namespace中,形成逻辑上分组的不同项目.小组或用户组,便于不同的分组在共享使用 ...

  6. jxl库介绍

    jxl是个韩国人开发的java中操作excel的库(棒子国思密达) 相对于另一个java excel库poi来说,jxl具有小巧和使用简单等优点. File uploadedFile = new Fi ...

  7. Day 6 编码的进阶

     https://blog.csdn.net/Deft_MKJing/article/details/79460485 a.ascii码:8位表示一个字符,共可以表示2**8个(即256)字符 ,   ...

  8. Flask从入门到精通之flask程序入门

    初始化 所有Flask程序都必须创建一个程序实例,Web服务器使用一种名为Web服务器网关接口的的协议(WSGI),把接收自客户端的所有请求转发给这个对象处理.程序实例是Flask类的对象,使用下面代 ...

  9. Java Web之Servlet中response、request乱码问题解决

    Java Web之Servlet中response.request乱码问题解决   一.request请求参数出现的乱码问题 get请求: get请求的参数是在url后面提交过来的,也就是在请求行中, ...

  10. (原创)确保JAVA线程安全的4种常用方法

    在Java中可以有很多方法来保证线程安全,比如使用同步方法.同步块,使用原子类(atomic concurrent classes),实现并发锁,使用volatile关键字,使用不变类和线程安全类. ...