亚洲丰满熟女一区二区哦,亚洲18在线天美,亚洲成人高清在线

Restatement: 性能,定w��,负蝲,以及(qi��ng)压力��试

切尔斯基 — Mon, 26 Jul 2010 14:44:00 GMT

摘要: 其实所有的一切都只是几个因素的相互作�? 互�ؓ(f��)函数:

* �q�发�?数据�?
* 机器配置
* 单个��h��处理速度
* �E�_��q�行旉��
阅读全文

切尔斯基 2010-07-26 22:44 发表评论

切尔斯基 — Sun, 11 Jul 2010 14:18:00 GMT

IoC避不开的一个问题是如何处理应用�E�序的模块化, 因�ؓ(f��)IoC通常针对单个对象提供�?ji��n)良好的支�? 比如依赖��理,生命周期��理,部��v旉��|�甚臌��行时配置, 但往往一�l�内聚的互相协作的对象才构成应用�E�序基本的构建块. �q�组内聚对象间的协作关系是实现细�? 包括单个对象的构造函数和属性也�? 如果把这些暴露出�? 固然可以提高灉|��? 但是以最后的部��v阶段的复杂性以�?qi��ng)难以维护性�ؓ(f��)代�h(hu��n)�? 以配�|�文件�ؓ(f��)�? 可能需要在包含�?ji��n)几百个对象定义的配�|�文件中, ��L��动七八个属�? 变换五六个实�? 来适应不同的运行环�?br />
两个解决�q�个问题的方�? 一个是IoC框架提供模块化支�? 以及(qi��ng)部��v时支�? 一�U�是IoC框架应该避开�q�个问题, �q�本不属于IoC的范�? 应该交由专门的模块化框架如OSGi去解�?br />
在我相对熟�?zh��n)�的IoC框架�? Autofac试图提供"Module"�q�个概念来解决一部分问题. 对象间的协作可以隐藏在Module对象�? 一些必要的参数可以通过Module的属性设�|�进�? 而对使用�?寚w��|�只需暴露Module对象卛_��. �q�是以��q��z�L��ؓ(f��)代�h(hu��n)�? 替换某个内部对象的实现变成了(ji��n)Hack, 需要通过团队成员间的交流, 源代码集体所有来完成. 如果在Module对象内部枚�D扩展点的所有实�? 通过部��v时的参数来挑(xi��)选某个实�? 则会(x��)引入不必要的依赖, 更是得不偿失

个�h們֐�于第二种. 谁介�l�一�?Net�q�_��上的模块化框�?

所以相对聚焦一�? �?ji��n)解一下IoC框架在处理单个对象的时候所需要解决的问题

1, 部��v时如何方便的支持产品环境和测试环�? 卛_��不同环境下提供同一接口的不同实�?/h3>
�q�个甚至不需要IoC框架解决, 应用开发�h员在不同环境下提供不同配�|�文件即�? 那问题变成了(ji��n)

2, 如何��化配�|? 使得框架能够自动识别依赖

�q�个被称为Auto wire, 一般根据构造函数参数或属性的�c�d��来匹配依�? 那么带来的一个问题是

3, �q�行时如何动态切换同一个类型的不同实现, 比如用户选择�?ji��n)只��L��? 那么所有的Save操作都应被忽略不做�Q何事, 可以通过提供Save操作的Dummy版本来实�?/h3>
Auto wire�Ҏ(gu��)��c�d��匚w��有一定的局限�? 卛_��一个运行时环境有某个接口的多个实现怎么�? �q�里实际上有一个概�? 卛_��象的标识. 而类型只是标识的一�U? 另外一�U�适应性更�q�的标识是字�W�串. 有的框架�U�C��为ID, 有的�U�C��为name. 标识的重要特征就是在一个运行时环境里是唯一�? �c�d��标识因此也可以�{化�ؓ(f��)字符串标�? 取其全名��可以了(ji��n). (标识本��n是个概念, 可以用一个类来表�? 然后有不同的子类实现, 如基于类型的标识, ��Z��字符串的标识�{?

因此��有两种�{�略: 一�U�是允许�q�行时某个对象可以重新声明依赖的标识, 一�U�是�q�行时用同样标识的对象替换原先的实现. 后一�U�相对容易实��C��? 按注册顺序Last One Win��可以了(ji��n)

如果同一个类型的多个实例需要同时存�? 则部�|�时�l�予不同的ID卛_��. 然而另一个问题是

4, 多用��L(f��ng)��境下, 如何保证每个用户都有自己的对象组合而不互相�q�扰

IoC框架一般被�U�Cؓ(f��)IoC容器. 容器�q�个概念是对�q�行时环�?上下文的��装, 提供�?ji��n)一�l�基��设施, 以及(qi��ng)�q�行时所需的各�U�服�? 最重要的一�Ҏ(gu��)��, 它提供了(ji��n)在同一个进�E�里的隔��? 多用��L(f��ng)��境下, 可以创徏多个彼此独立的容�? 每个容器负责创徏所需的对�? 提供所需的服�? 彼此之间互不影响. 于是带来新的问题,

5, 对象的生命周期如何管�?/h3>
很显�? 有的对象概念上属于全局, 在程序运行过�E�中不能创徏多个实例, 也不能被销�? 有的跟用��L(f��ng)��一�ơ操�? 一�ơ请求相兌��, 同一�ơ请求中可以复用同一个对象实�? 但不同请求必��d��Z��同的对象实例. 有的则每�ơ需要访问它的时候都得创��Z��个新的实�? 容器必须提供基础设施, 让应用开发者可以指定对象的生命周期, �q�在周期�l�束旉��毁所有该销毁的对象. C# 提供�?Dispose 机制, 可供利用

其它的一些问�? 比如:

如何支持composite, decorator, proxy模式

如何方便用户注册, 包括自动注册

如何注入容器创徏的对象到用户手工创徏的对象中

如何与现有各�U�框枉��?/li>
如何避免循环引用: Constructor/property dependencies

Open generics injection

List injection Unregistered resolution

Auto-mocking

Startable

Strongly-typed Activation Events

Adding to an Existing Container

切尔斯基 2010-07-11 22:18 发表评论

切尔斯基 — Tue, 06 Jul 2010 15:15:00 GMT

Log 问题�?br />

如何不带来额外的效率损失
如何在程序运行出错时记录��可能多的信�?/li>
如何方便查找特定条�g的错�?/li>
如何横切的添加通用信息

如何不带来额外的效率损失

在之前接触的一个大型��品中见过散布着如下代码:

if (Log.Level > DEBUG) {
logger.write(some_method_to_build_the_log_string());
}

问�ؓ(f��)什么不在logger.write()里面判断日志�U�别, �q�样外面写�v来就很清�?br />
logger.write(some_method_to_build_the_log_string());

�{�曰效率问题, 后面一�U�写法会(x��)��D��无论日志最�l�有没有被写�? some_method_to_build_the_log_string()都会(x��)先行被调用求�? 而这可能很耗时...

一个解军_��法是额外的中间层: 能够延时求值的东东, 比如函数指针, 函数对象, 总而言之能够保存当前上下文而在��来某个时刻能够回调的东�? 诸如 C# 里的 Func

�q�样logger.write()的签名就�?void write(string info) 变成 void write(Func getInfo);

带来的一个问题是啥时真正写入, 一般可以是用户事务�l�束的时�? 其实当�ƈ发用户量大时, 表面上�g时求值可以不�ȝ��当前��h��的处理速度, 但当求值发生时, 可能�?x��)阻��其它请求的处�? 所以Func代替string效率上的优势�q�不明显, 更多的是代码上的��z?br />
(Func的另一个好处是把IO操作推迟�? 当时没有IO操作�Q�所有IO操作都被推到完成用户��h��之后再进行，��其是如果因为其它原因需要把Log存到数据库的时�? 当然不用Func, 直接用string也可以实现推�q�IO操作, 只需要把string都缓存�v�? 找个时机再写出来卛_��)

然而Func更大的意义是解决�W�二个问�?

如何在程序运行出错时记录��可能多的信�?/h3>
通常最�l�写到日志里的信�? 都是�Ҏ(gu��)��配置的��别决定的, 比如配置�?ji��n)日志��别是INFO, 那么无论��Z��出错, DEBUG�U�别的信息都不会(x��)被写到日志里. 然而现实情冉|��, 一旦出�? 我们可能需要DEBUG信息来定位问�? 于是我们不得不把日志�U�别讄��成DEBUG,然后重新�q�行应用,企图复现. 但运行环境的差异使得复现像撞大运, 可遇不可�? 于是在案发第一现场��p��录所有线索变得极��h��?br />
�Ҏ(gu��)��志即时求值和写入是很隑֮�现这个目的的, 而�g时求�? 回调, �q�层额外的间接则��其变得轻而易�?

我们要做的就是保存所有写日志的回�? 直到写入的那一�? �Ҏ(gu��)��某种规则, 从中�?xi��)选部分回调真正去执行; 规则可以是正常情况下�Ҏ(gu��)��日志�U�别讄��, 出错的时候则全部写入, �{�等

(�q�里的目的就是避免日志过于简略或者繁�? 鼓励大家多写日志而不必担�?j��)运行时效�?

如何方便查找特定条�g的错�?/h3>
查找是数据库的强��? 把日志结构化, 存到数据库里卛_��. �q�会(x��)带来事务的问�? 写日志是否和用户正常的业务放在一个事务里?

如何横切的添加通用信息

如果把log建模��Z��个对�? 可以应用builder模式, pipeline�{? 让log对象依次通过一堆builder�l�成的pipeline, 每个builder负责�l�log增砖�ȝ��, 最后出来的��是一个包含了(ji��n)各种彼此独立的信息的对象

切尔斯基 2010-07-06 23:15 发表评论

切尔斯基 — Mon, 28 Jun 2010 14:01:00 GMT

谁来做Validation
何时做Validation
如何表达错误
如何传递错�?/li>
如何兌��错误到发生错误的对象, ��其是对象图中非Root对象

�q�里的Validation指的是对�q�入到系�l�中的业务数据的校验(不包括Web应用中页面数据在��览器端的验�?

谁来做Validation

数据的有效性不是自�w�所能决定的, 而是使用它的场景(Context)军_��? 因此, 每个Context应该有自��q��Validation逻辑.

一个例�? 个�h信息�D�缺, 比如婚姻状况没填, 但联�p�d��址�?sh��)子邮�g�{�信息完�? 那么�q�个个�h信息到底是合法还是非�? 如果你的应用是个�E�务相关的应用必��ȝ��道婚�ȝ��况则数据是非法的, 如果你的应用是CRM�pȝ��有客��L(f��ng)��联系方式卛_��而婚�ȝ��冉|��可选的则数据就是合法的. 问题是你的应用是�E�务应用但同时支持客户关�pȝ��? 那这�D�|��据到底合法非�? �E�务应用只是�q�底的时候才有�h�? 而客户关�pȝ��理系�l�随旉��有�h�? 假设数据是通过��面提交�? 那这�Ҏ(gu��)��据到底该拒绝�q�是接受?

何时做Validation

通常有几个时�? 对象被创建出�? 对象状态改�? 以及(qi��ng)对象被持久化. 企业应用中同一份数据一般至��有两种存在形式: 在数据库中的持久化状�? 以及(qi��ng)在内存�(sh��)��以编�E�语�a�定义的对象�Ş式存�? 那么几个时机:

手动创徏对象, ��是应在构造函��C��做验�?/li>
框架帮忙创徏对象, 比如从页面Form�l�定到Server端的对象�? 可以在绑定完成的那一��d��校验
存到数据库里那一�?/li>

�q�就带来一个问�? 可能要在三个地方做大体相同的验证, 如何复用验证规则?
另一个问题是: 在引入ORM的应用中, �~�程语言写好的验证逻辑同样可以应用在持久化到数据库的那一�? 那在SQL/DDL语句中定义的�U�束是否�q�必�?

如何表达错误

有两个约�?

要提供易于用户和支持人员理解的错误信�?/li>
要提供尽可能丰富的信�?/li>

常见的手�D�|��用字�W�串或者错误代�?ID, �q�是不work�? 因�ؓ(f��)它们合�ƈ�?ji��n)错误本�w�和错误的表�C?

出错的地方可能距��需要展现错误的地方很远, 或者有多种展现错误的界�? 或者有很多昄��斚w��的需�? 比如支持国际�? 报错的地�Ҏ(gu��)��没有能力也不需要知道错误是如何被展�C�的, 它要做的是尽可能报告关于错误的详�l�信�? 包括�q�反�?ji��n)什么规�? 出错的字�D? 实际的值和期待的值等, 字符串和错误代码/ID是没有如此丰富的表达能力�?br />
我们可以用对象来表达错误信息, 对象的类型可以表�C�错误的�c�d��, 对象的属�?字段可以携带各种与错误类型相关的数据. 然后在需要展现给用户的那一�? 再把对象��译成针寚w��个界面的昄��, 比如可以做国际化, 或者提供给�E�序员更技术化的描�q? 而在错误信息需要被昄��之前, 错误在系�l�中的传�? 都是以对象的形式�q�行�?..

听�v来跟异常Exception很像?

几个反例�?Net�q�_��上的异常, 比如KeyNotFoundException, 它就不告诉你那个找不到的Key是啥, �q�有Index��界, ��׃��告诉你index的值是多少, �q�有数据库连接超时或者Transaction Timeout,��L��不告诉你它等�?ji��n)多久超时�? 让你搞不清楚是你的超时时间设的太短还是根本你的设�|�就没生�?br />

如何传递错�?/h3>

攉��参数, 输出参数, Thread Local, 或者抛出异常然后合适的层次捕获

如何兌��错误到发生错误的对象�? ��其是对象图中非Root对象

�l�错误一个Key, �q�个Key应该能表�C�出错的对象在对象图中的位置, 比如Key可以是字�D�名�U�C��间用"."分隔, �U�联��h��的字�W�串. 注意�q�种形式的key应该是在调用验证逻辑的地方组装�v来的, 而不应该是验证逻辑本��n, 因�ؓ(f��)验证逻辑通常�q�不知道自己验证的这个对象在父对象中的字�D�名�U?br />

一个额外的话题

很多验证框架采用�?ji��n)基于Attribute/Anontation的方�? �q�样当一个对象在不同的Context有不同的验证逻辑时就�?x��)很�U�结, 因�ؓ(f��)它是以��R入的方式写到对象的定义中�? �q�恰��C��另一个角度说明了(ji��n)对象是不应该跨Context复用�? DCI才是王道

切尔斯基 2010-06-28 22:01 发表评论

切尔斯基 — Sun, 20 Jun 2010 15:37:00 GMT

同一个Feature的代码要攑֜�一�?IDE里单独的一个工�E? 或者工�E�里单独的一个文件夹), �q�些代码要么全有要么全无�? 它们合作完成一个Feature, 如果用户不再需要这个Feature�? 可以把它们整个的痛快的删�? 不会(x��)留下谁也用不到的代码成�ؓ(f��)�pȝ��的垃�? 如果想看一个Feature是如何实现的, 那所有相关代码都在一�? 不需要在庞大的代码库中蟩来蟩�?

那么理想的情况就�? 你看看源代码�?w��i)里所有工�E�文件的名字, 或者文件夹的名�? ��q��道系�l�提供了(ji��n)哪些功能, 它可以跟你的需求描�q�对应�v�? 无论用User Story�q�是Use Case, 都可以用它们的名字作为工�E�名或者文件夹的名�? 方便�l�护

��行的MVC框架�~�省的物理文件组�l��ƈ不是�q�样�? Controller, Model, View分别在不同的文�g多w��? ASP.Net MVC提供�?ji��n)VirtualPathProvider以及(qi��ng)ViewEngine, 可以让我们把一个Feature的Controller/Model/View�l�统打包��C��个project或者文件夹而运行时依然能够扑ֈ�对应的action和view, �q�是我们正在利用的特�?br />

�q�种代码�l�织方式�Ҏ(gu��)��构的影响是什�?

�q�基本会(x��)��D��Z��插�g/扩展点的体系�l�构. 攑ֈ�更大��度�? ��是SOA. SOA才是王道. �q�个词太大了(ji��n), �q�是先聚焦到一个进�E�的应用....

1. UI如何聚合? 最�l�用��L(f��ng)��到的UI, 是一个聚合的�l�果, 可能来自�pȝ��的不同部�? 解决�q�个问题的扩展点技术有客户端的Ajax, 或者服务端的RenderAction. (问题: Css应该如何处理? 不同部分的显�C�顺�? 布局如何��定?)
2. Feature如何沟�? Feature之间不可能一点依赖没�? 比如可能�?x��)用到相同的数�? 相同的业务逻辑. 解决�q�个问题的方法有Bounded Context, Context Mapping, DCI...都是一回事
3. 数据库如何划�? 不同的Feature使用自己的独立定义的数据�? 做映��和同步, 也是同样的方�?br /> 4. 如何把这些Feature�l�装在一�? Java�q�_��有OSGi, .Net目前没有看到跟OSGi�c�M��的方�? 基本是注册或动态发现的路子, 遵��@开闭原�?..

切尔斯基 2010-06-20 23:37 发表评论

切尔斯基 — Mon, 28 Dec 2009 15:49:00 GMT

摘要: 假设我们必须处理对象的存�? 加蝲, 和查�? 性能和引用完整性的�U�束, �l�接口的实现带来�?ji��n)以下问�?

1. 加蝲根对象时如何避免加蝲大半个数据库
2. 存储时如何更新整个对象图
3. 存储时如何高效的更新整个对象�?
4. 何时同步对象的内存状态和持久存储状�?
5. 如何��保在出错时保持对象内存状态和持久存储状态之间的一致�?
6. 如何保证引用的唯一性以避免可能的更新冲�H?

�Ҏ(gu��)��能的精益求�_? 又促(j��)使�h们解��x(ch��ng)��多的�l�节问题:

7. N+1查询问题
8. 分离查询模型和存储模�?
9. ��量减少查询语句

�q�些问题的解��x(ch��ng)��案又�?x��)带来新的问�? 阅读全文

切尔斯基 2009-12-28 23:49 发表评论

Web开发问题域

切尔斯基 — Sun, 06 Dec 2009 10:10:00 GMT

摘要: 其它所有问题都是这个问题的某种解决�Ҏ(gu��)��引入�?nbsp; 阅读全文

切尔斯基 2009-12-06 18:10 发表评论

切尔斯基 — Mon, 28 Sep 2009 15:05:00 GMT

摘要: 我们都知道我们无法证明��Y件已�l�没有Bug, 用�L普尔的话说就�? �U�学理论都是假说, 爱因斯坦的竞争理��明对牛顿理论的即使如��L(f��ng)��星发现般严格的检验都不能��保其正��? 卛_��白天鹅的一千次观察都不��以断言黑天鹅的不存�?nbsp; 阅读全文

切尔斯基 2009-09-28 23:05 发表评论

跨团队的持箋(hu��)集成: 几个基本矛盾

切尔斯基 — Tue, 07 Jul 2009 14:20:00 GMT

摘要: 1. �q�行开�?�?集成之间的矛�?
2. ��Z��文本的合�q?�?��Z��语意的逻辑之间的矛�?
3. 依赖的稳定�?�?依赖自��n的演�q�升�U?之间的矛�?nbsp; 阅读全文

切尔斯基 2009-07-07 22:20 发表评论

切尔斯基 — Sun, 05 Jul 2009 08:43:00 GMT

摘要: 试错法广泛的应用于自然科学领�? �q�代开发就是试错法在��Y件开发过�E�方面的应用阅读全文

切尔斯基 2009-07-05 16:43 发表评论

切尔斯基 — Thu, 04 Jun 2009 15:43:00 GMT

摘要: 机器环境虽然各有各的不同, 但依然有可能创徏一�?环境无关的环�? 阅读全文

切尔斯基 2009-06-04 23:43 发表评论

ThrowawayObject :在Java中实现Rails里的Flash Message

切尔斯基 — Mon, 06 Oct 2008 14:26:00 GMT

摘要: �q�应该是一个概�? 你放一个东西到Session�? 但你只能获得一��? 下次它就不存在了(ji��n). 用类来表�C�概�? 我们只需在生成对象时把自己加到Session�? 在提供的获取函数中把自己从Session中移�? 那么�W�二�ơ你��得不到它了(ji��n). 阅读全文

切尔斯基 2008-10-06 22:26 发表评论

敏捷质疑: �l�对�~�程, 代码集体所有权

切尔斯基 — Sun, 27 Jul 2008 14:25:00 GMT

摘要: Q: �l�对�~�程、责��d��享，完全是胡��_(d��)��代码找不��C��者，开发�h员哪里会(x��)有责��d��(j��)!

Q: 我干嘛要把辛辛苦苦很多年�U�篏的经验白白告诉别�? 我喜�Ƣ不可替代的感觉.

Q: 有些老手不喜�Ƣ结�? 觉得��C�h不劳而获对他们不�? 不情�? 怎么�?

Q: 那互不相让怎么�?

Q: 我想修改某段代码, ��x(ch��ng)��原作者了(ji��n)解一下思�\, 可根本不知道是谁阅读全文

切尔斯基 2008-07-27 22:25 发表评论

敏捷质疑: 持箋(hu��)集成

切尔斯基 — Sun, 20 Jul 2008 13:55:00 GMT

摘要: Q: 我的产品是电(sh��)信��的设�? 几百人分成几十个��目�l�在开�? 各个��目�l�进度不�l�一, 如何集成?
Q: CI的愿景是好的, 但我们这里根本不可能, 我们的��品需要复杂的�q�行环境, �q�行旉��要�h工干�? 怎么��? 阅读全文

切尔斯基 2008-07-20 21:55 发表评论

敏捷质疑: TDD

切尔斯基 — Sun, 13 Jul 2008 13:40:00 GMT

摘要: Q: ��Z��么通过单元��试发现�?Bug 很少 ?

Q: 你们说验收测试应该由客户来编�? 可在我们�q�里�Ҏ(gu��)��不可�?
阅读全文

切尔斯基 2008-07-13 21:40 发表评论

切尔斯基 — Mon, 07 Jul 2008 14:14:00 GMT

摘要: Working Effectively with Legacy Code 阅读全文

切尔斯基 2008-07-07 22:14 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(6) : Scaling up

切尔斯基 — Sat, 12 Apr 2008 15:09:00 GMT

摘要: 挖掘?sh��)?CI server 的能�?nbsp; 阅读全文

切尔斯基 2008-04-12 23:09 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(5) : Refactor your configuration file

切尔斯基 — Thu, 06 Mar 2008 15:26:00 GMT

摘要: 希望�q�能帮助让你的CruiseControl的配�|�文件更�Ҏ(gu��)��l�护. 我相信持�l�集成应该很��? 即�ɘq�会(x��)让我�׃��. 阅读全文

切尔斯基 2008-03-06 23:26 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(4) : Bootstrap with a Bootstrapper

切尔斯基 — Sun, 17 Feb 2008 08:09:00 GMT

摘要: �H�然之间多了(ji��n)点对你的构徏破坏者同事们的容�? 你盯着那个 build.xml, 它依然显�C�在你的IDE里面. ��是它了(ji��n), 文�g�l�尾处有一个多余的大大�?'#' �?nbsp; 阅读全文

切尔斯基 2008-02-17 16:09 发表评论

切尔斯基 — Sat, 12 Jan 2008 14:03:00 GMT

摘要: Ant脚本如何适应大规模的��目, Large Scale 的项目意味着: 目录较多, 层次较深依赖较多, 构徏脚本依赖的第三方Ant Task, ��目依赖的第三方库等 ��试较多, 构徏旉��反馈周期较长需要在不同操作�pȝ��上运�?需要在不同团队成员的机器上�q�行 Ant脚本较长如何�l�护阅读全文

切尔斯基 2008-01-12 22:03 发表评论

切尔斯基 — Sat, 05 Jan 2008 10:29:00 GMT

摘要: "拥抱变化" 是敏��L(f��ng)��态度之一, CruiseControl 正是来实证这�U�态度的作�? 多种�c�d��?变化"都会(x��)触发CruiseControl的一�ơ构��E?

我们知道CruiseControl能根据源代码的变化来调度一�ơ构�? 但你知道CruiseControl支持多少�U�调度模式吗?
阅读全文

切尔斯基 2008-01-05 18:29 发表评论

Ivy: 搭徏本地仓库

切尔斯基 — Sun, 23 Dec 2007 15:22:00 GMT

基本�? 搭徏一个本地的ivy仓库, 包含两�g独立的事�? 搭徏仓库本��n, 和配�|�如何��用这个仓�?/p>

仓库本��n

Ivy的Repository是由一个个Module�l�成�? Module则包含自�w�的artifacts(通常是jar文�g)和描�q�文�?

Module的描�q�文件是ivy的核�? 里面记录�?ji��n)这个module包含哪些artifacts和这个module对其它module的依�? �q�样��p��摸瓜, 牵出所有的依赖�?/p>

<ivy-module version="1.0">

<info organisation="your.company" module="your.project" />

<configurations>

<conf name="release" />

<conf name="testing" extends="release" />

configurations>

<publications>

<artifact name="common" />

<artifact name="client" />

<artifact name="server" />

publications>

<dependencies defaultconf="release->default">

<dependency name="ant" rev="1.7.0" />

<dependency name="antlr" rev="2.7.6" />

<dependency name="xstream" rev="1.2.2" />

dependencies>

ivy-module>

仓库在文件系�l�上的结�? 可以�׃��自己军_��, 只要在��用这个仓库时, 把它的布局信息用配�|�文件描�q�出来告诉��用者就可以�?/p>

如何使用�q�个仓库

基本�? 我们借助 Ant 来��用ivy, 那么我们需要告�?Ant 一些repository相关的信�?: Where is the local repository, and How it looks like

<property name="ivy.local.default.root" location="/your/local/ivy/repository/folder" />

<ivy:settings id="ivy.instance" file="${ivy.local.default.root}/ivy.repository.settings.xml" />

那个ivy.repository.settings.xml��是来描�q�repository的布局�?

<ivysettings>

<settings defaultResolver="local" />

<resolvers>

<filesystem name="local" checkmodified="true">

<artifact pattern="${ivy.local.default.root}/[module]/[artifact].jar" />

<artifact pattern="${ivy.local.default.root}/[module]/[artifact]-[revision].jar" />

<ivy pattern="${ivy.local.default.root}/[module]/[module]-dependencies.xml" />

filesystem>

resolvers>

ivysettings>

然后你告�?Ant 你的��目的依赖和artifacts, Ant��可以帮你获得依赖和发布artifacts�?/p>

<property name="ivy.dep.file" location="${basedir}/my-module-dependencies.xml" />

<ivy:retrieve />

resolver="local" pubrevision="1.0" overwrite="true" forcedeliver="true">

pattern="${build.dir}/dist/[artifact].[ext]" />

pattern="${basedir}/my-module-dependencies.xml" />

那个my-module-dependencies.xml��是你的��目的依赖描�q�符, 事实上ivy也将它看作一个module,和repository里的module一视同�? 参考最开始的例子.

Ivy 如何解决我们的问�?/h3>
Q: 我的��目在开发环境和产品环境有不同的依赖, 怎么�?

A: Ivy 提供�?configuration" 的概�? 在那�?font color="#2a00ff" size="2">my-module-dependencies.xml�? 你可以�ؓ(f��)你的module定义development和product两种配置, 可以指定每个artifact隶属于哪个配�|? 指定每个dependency属于哪个配置. 注意�q�是一个递归定义, 因�ؓ(f��)你依赖的module也有可能定义�?ji��n)多个配�|? 你需要指定你依赖于依赖的哪个配置,所以这是一个映��? 比如你的产品在运行环境中依赖于spring的��品环�?

<ivy-module version="1.0">

<info organisation="your.company" module="your.project" />

<configurations>

<conf name="product" />

<conf name="development" />

configurations>

<dependencies defaultconf="product->default">

<dependency name="spring" rev="1.7.0" conf="release->product" />

<dependency name="antlr" rev="2.7.6" conf="development->debug" />

<dependency name="junit" rev="4.4" conf="development->release" />

dependencies>

ivy-module>

ivy 的各个Ant task允许你指定在哪个 configuration 上操�? 如�ؓ(f��)product配置取得所有依�?

<ivy:retrieve conf="product" />

Q: 我的��目在开发环境和产品环境有太多相同的依赖, 事实上开发环境包含��品环境所有的依赖, 额外再加上junit,jmock�{? 如何消除重复的依赖描�q�呢?

A: Ivy 提供�?ji��n)配�|�之�?"extends" , 也就�?�l�承"的概�? 你可以让开发环境的配置�l�承自��品环境的配置, �q�样��可以复用��品环境的配置

<configurations>

<conf name="product" />

<conf name="development" extends="product" />

configurations>

Q: �~�省ivy��L��从缓存�(sh��)��d��配置, �q�样我的依赖配置更新后却得不到反�? 怎么��止从cache中读取配�|?

A: <filesystem name="local" checkmodified="true">

切尔斯基 2007-12-23 23:22 发表评论

RegexBuilder: A DSL for creating regular expression, Request For Comment

切尔斯基 — Sun, 16 Dec 2007 13:36:00 GMT

class IP

include RegexpBuilder

def less_than_255

_0_to_100 = repeat(digit(), 1, 2)

_100_to_200 = "1" + _0_to_100

_200_to_250 = "2" + either("0-4") + either("0-9")

_250_to_255 = "25" + either("0-4")

one_of(_0_to_100, _100_to_200, _200_to_250, _250_to_255)

end

def initialize

@ip_pattern = repeat(group(group(less_than_255()) + literal(".")), 3) + group(less_than_255())

end

attr_reader :ip_pattern

end

以上是应�?RegexBuilder 书写正则表达式例�?IP 地址的一个例�?

RegexBuilder 的目的是试图增强正则表达式的可读性和可维护�?/p>

RegexBuilder �q�不是用来取�?Regexp �? 而是辅助�~�写 Regexp 构造函数需要的�W�一个参�?

RegexBuilder 包含�?ji��n)正则表辑ּ��W�号(Anchors/Character Classes/Repetition/Alternation/Grouping)�?API 的一一对应, �?one_of => |, either => [], at_least_one => + �{?

RegexBuilder �q�包含了(ji��n)可以直接使用的一�l�常见的正则表达�? �?IP 地址, Email 地址�{?

RegexBuilder 使用�?ji��n)Java风格的API而不是Ruby风格的API, �?repeat(3, digit()) 而不�?digit(:repeat => 3)

RegexBuilder �q�没实现Substitutions, �?\0, \1, \& �{?

安装下蝲

gem install regexbuilder

http://roll-stone.googlecode.com/svn/trunk/RegexBuilder/

切尔斯基 2007-12-16 21:36 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(3) : Configuring CruiseControl the CruiseControl way

切尔斯基 — Thu, 08 Nov 2007 16:57:00 GMT

摘要: 我想�q�个实践的核�?j��)�?

1, ��CruiseControl的配�|�文件check in到版本控制系�l�中, 以解��x(ch��ng)��外损坏的问题

2, 使用专门�?project"来自动更新配�|? 以解��x(ch��ng)��ơ需要有��Z��门登录到build server上去更新的瓶颈问�?

3, 复用但不依赖于CruiseControl寚w��|�文件的validation, 以同时获�?"��L��有问题的配置文�g被应用到build server的能�? �?"�q�速获知有人check in�?ji��n)有问题的配�|�的能力"
阅读全文

切尔斯基 2007-11-09 00:57 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(2) : Keep your dependencies to yourself

切尔斯基 — Mon, 15 Oct 2007 15:13:00 GMT

摘要: 让你的项目包含自��q��依赖. (让你的构��本显式的引用包含在项目内部的依赖, 而不是隐式的让构建工��h��包含那些依赖) 阅读全文

切尔斯基 2007-10-15 23:13 发表评论

CruiseControl Enterprise 最�?j��ng)_��?(1) : Publish with a Publisher

切尔斯基 — Thu, 11 Oct 2007 16:02:00 GMT

摘要: 我想�q�个实践的好处就�?

1. 开发者每�ơ在自己机器上构建时不需要发�? 省时�?
2. CruiseControl使用跟开发者相同的构徏脚本, 减少�?ji��n)开发者构建成功而CruiseControl构徏��p�|的概�? 省调试时�?
3. CruiseControl�q�行Publisher时开发者可以��l�工作了(ji��n), 提高?sh��)��(ji��n)�ƈ发�? �q�是省时�?nbsp; 阅读全文

切尔斯基 2007-10-12 00:02 发表评论

jadestone enhanced sax handlers 2.0 发布

切尔斯基 — Tue, 12 Jun 2007 22:35:00 GMT

介绍: Enhanced SAX Handler :�?DOM �q�简单的 SAX Handler

下蝲: http://code.google.com/p/jade-stone-suite/downloads/list

�C�Z��: http://jade-stone-suite.googlecode.com/svn/trunk/JS.ESax/test/jade/stone/esax/sample/FACupHandler.java

切尔斯基 2007-06-13 06:35 发表评论

[导入]��单至�?qi��ng)的AOP和IOC

切尔斯基 — Sun, 03 Jun 2007 21:27:00 GMT

随处可见的AOP和IOC
文章来源:http://blog.csdn.net/wuliaoshengji/archive/2006/03/26/639007.aspx

切尔斯基 2007-06-04 05:27 发表评论

[导入]Configuration Pattern : 配置模式

切尔斯基 — Sun, 03 Jun 2007 21:27:00 GMT

IoC? Dependency Injection? No, Configuration Pattern!
文章来源:http://blog.csdn.net/wuliaoshengji/archive/2007/05/12/1605465.aspx

切尔斯基 2007-06-04 05:27 发表评论

[导入]Axis : �q�去前的备忘

切尔斯基 — Sun, 03 Jun 2007 21:26:00 GMT

simple axis and axis memo
文章来源:http://blog.csdn.net/platformlib/archive/2007/03/09/1525377.aspx

切尔斯基 2007-06-04 05:26 发表评论

Enhanced SAX Handler :�?DOM �q�简单的 SAX Handler

切尔斯基 — Mon, 21 May 2007 21:39:00 GMT

摘要: �q�实际上反映�?ji��n)目前SAX Handler接口定义中缺��q��三种能力: Stoppable, Subscribable, �?Reportable; 最�l? ESAX �?原始�?SAX Handler 补��?可中止的能力, 可订阅的能力, 可汇报的能力, 使得比原始的SAX Handler更快, 比DOM接口更简�? 更易于编�E?nbsp; 阅读全文

切尔斯基 2007-05-22 05:39 发表评论

emma : ��试覆盖率低于指标的时候让Ant Build自动��p�|

切尔斯基 — Mon, 21 May 2007 20:10:00 GMT

摘要: emma : ��试覆盖率低于指标的时候让Ant Build自动��p�| 阅读全文

切尔斯基 2007-05-22 04:10 发表评论

RegexBuilder: A DSL for creating regular expression

切尔斯基 — Fri, 24 Nov 2006 09:24:00 GMT

/* 最新版本请参见 http://www.tkk7.com/chelsea/archive/2007/12/16/168118.html */

找了(ji��n)半天也没有找到创建正则表辑ּ�的DSL, 如果你知道请告诉�?/p>

只好先自己写来应急了(ji��n): http://roll-stone.googlecode.com/svn/trunk/RegexBuilder/

用�v来类��D��?

require 'regex_builder'

class RegexBuilderTest < Test::Unit::TestCase

def setup

@rb = RegexBuilder. new

end

def test_ruby_start_with_r

pattern = @rb .start_with( "r" ).to_s

assert "ruby" .match(pattern)

end

def test_ruby_start_with_r_and_end_with_y

pattern = @rb .start_with( "r" ).any().end_with( "y" ).to_s

assert "ruby" .match(pattern)

end

�q�有一�?RegexUtil 来直接提供常用的正则表达�? 如email地址,ip地址�{?惛_��啥都可以往里加

require 'regex_util'

class RegexUtilTest < Test::Unit::TestCase

def test_192_168_0_3_is_a_valid_ip_address

assert "192.168.0.3" .match(RegexUtil::IP_Pattern)

end

def test_500_312_0_3_is_not_a_valid_ip_address

assert_nil( "500.312.0.3" .match(RegexUtil::IP_Pattern) )

end

def test_chelsea_at_gmail_dot_com_is_a_valid_email_address

assert "chelsea@gmail.com" .match(RegexUtil::Email_Pattern)

end

def test_www_dot_google_dot_com_is_not_a_valid_email_address

assert_nil( "www.google.com" .match(RegexUtil::Email_Pattern) )

end

很多功能都还没提�?希望有�h一��h��?/span>

切尔斯基 2006-11-24 17:24 发表评论

Get Mac Address in Java

切尔斯基 — Wed, 15 Nov 2006 13:34:00 GMT

(t��ng)

基本目前没有�U�Java的API来访问硬件信�?以后或许�?x��)有�?谁知�?/font>

对Mac地址的需求部分来自UUID, 另有很大部分来自License Management,�q�真是一件无聊的事情,希望以后软�g卖的都是服务,不是拯��,省却�q�些�ȝ��(ch��)

(t��ng)

搜了(ji��n)一�?大体有三�c�解军_��?/font>

1, JDK6, JDK7 ?

详情可参见以下链�?/font> http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4691932

2, Shell Command

��是截取 ipconfig, (t��ng) ifconfig �{�命令的输出,然后解析

有一些开源项�?�? http://sourceforge.net/projects/javanetcomm/

�q�有很多代码片断,�? http://forum.java.sun.com/thread.jspa?threadID=61556, (t��ng)http://forum.java.sun.com/thread.jspa?threadID=245711

�q�种�Ҏ(gu��)��的缺�Ҏ(gu��)��有可能需要较高的�pȝ��权限才能执行�q�些命��o(h��)或者访问网卡硬�?/span>

3, JNI

也有开源项目包含了(ji��n)�q�个功能, �? http://jug.safehaus.org/

�~�点是不方便部��v

(t��ng)

然而不是所有的�|�络都是Ethernet, 对License Management来说Mac地址只是更�ؓ(f��)generic的概�?Host Id"的一�U�考量因素.

切尔斯基 2006-11-15 21:34 发表评论

WS-Security Practice

切尔斯基 — Thu, 31 Aug 2006 14:58:00 GMT

摘要: 阅读全文

切尔斯基 2006-08-31 22:58 发表评论

Essential Vitria BusinessWare

切尔斯基 — Sun, 28 May 2006 15:44:00 GMT

(t��ng)

对BusinessWare的��用告一�D�落�Q��ȝ��一�?/p>

(t��ng)

(t��ng)Vitria BusinessWare: 事�g与端�?/a>�Q�摘要：(x��)Vitria BusinessWare: Event and Port (t��ng) (t��ng) (t��ng) (t��ng)�Q�全文共3431字）(j��)——点�?a target="_blank" >此处阅读全文

(t��ng)

(t��ng)Vitria BusinessWare: Web Services�Q�摘要：(x��)BusinessWare既可以作为WebService客户端调用其它应用发布的WebService�Q�又可以作�ؓ(f��)WebService服务端，��整个流�E�模型发布�ؓ(f��)WebService�Q�供其它应用调用�Q�而这一切都可以在徏模工具BME中通过可视化的界面和向�?Wizard)来方便的完成 (t��ng) (t��ng) (t��ng) (t��ng)�Q�全文共6580字）(j��)——点�?a target="_blank" >此处阅读全文

(t��ng)

(t��ng)Vitria BusinessWare: �q�_��与��Y件�ȝ��Q?/a>摘要�Q�BusinessWare中SOA的成分越来越多�?t��ng) (t��ng) (t��ng) (t��ng)（全文�?205字）(j��)——点�?a target="_blank" >此处阅读全文

(t��ng)

(t��ng)Vitria BusinessWare: 存储与访问安�?/a>�Q�摘要：(x��)事实上，BusinessWare使用LDAP做�ؓ(f��)存储机制和安全机制�?t��ng) (t��ng) (t��ng) (t��ng)（全文�?2865字）(j��)——点�?a target="_blank" >此处阅读全文

(t��ng)

(t��ng)Beyond Workflow : An Introduction to Vitria BusinessWare�Q�摘要：(x��)BusinessWare是一�ƾ企业应用集�?EAI)软�g�Q�更加注重“系�l�”间的交互，传统工作��只是它的一��Ҏ(gu��)��通功能而已�Q? BusinessWare�q�是一�Ƒ֏�扩展的��Y�Ӟ��对于已知的未知的�pȝ��/协议�Q�它提供可单独配�|�的部�g(Connector/Proxy)来对接，因此�Q? BusinessWare可以看作企业现有�pȝ��间的�ȝ��/�E�控交换机�?t��ng) (t��ng) (t��ng) (t��ng)（全文�?505字）(j��)——点�?a target="_blank" >此处阅读全文
(t��ng)

切尔斯基 2006-05-28 23:44 发表评论

Essential Web Services: SOAP, WSDL, UDDI

切尔斯基 — Wed, 24 May 2006 16:07:00 GMT

摘要: 1. SOAP: �?Web Service 无关
2. WSDL: �?Runtime 无关
3. WSDL-SOAP Binding Style
4. UDDI�Q�与 WSDL 无关
5. UDDI�Q�与WSDL 有关
6. JAX-RPC: 首先是Java,其次才是RPC 阅读全文

切尔斯基 2006-05-25 00:07 发表评论

切尔斯基 — Wed, 22 Feb 2006 12:31:00 GMT

摘要: 阅读全文

切尔斯基 2006-02-22 20:31 发表评论

切尔斯基 — Thu, 12 Jan 2006 03:22:00 GMT

I. Questions before reading the specification, and Answers after reading the specification.

1. How the application call EIS functions to execute task or get data from EIS via JCA RA/Connector?

This is addressed by Common Client Interface, it involves "Interaction", "InteractionSpec", "Record" and other interfaces to do this work.
It is a Command Design Pattern

2. and on the other hand ?

This is addressed by Message Inflow Contract and EJB Client View Model together. It involves a MDB to dispatch invocation to application components.

3. CCI is a protocol between Resource Adaptor and EAI tools ?

You can understand it as this. CCI is so low level so that it is targeted primarily towards application development tools and EAI frameworks.

4. How the ActivationSpec get associated with MDB instance ?

At deployment time, it is the deployer to associate ActivationSpec instance with MDB.
At Runtime, it is the Resource Adapter to associate ActivationSpec instance with MDB insatnce while the application server calls endpointActivation(MessageEndpointFactory, ActivationSpec)

II. Notes

1. Overview

Roles: J2EE application, J2EE AppServer, Resource Adaptor, EIS, and their provider

The target audience for this specification includes:

EIS vendors and resource adapter providers
Messaging system vendors
Application server vendors and container providers
Enterprise application developers and system integrators
Enterprise tool and EAI vendors

The system-level contracts between an application server and an EIS are targeted towards EIS vendors (or resource adapter providers, if the two roles are different) and application server vendors. The CCI is targeted primarily towards enterprise tools and EAI vendors.

JCA is a standard between EIS and J2EE Application Server rather than a standard between EIS and J2EE Application, for the standard defined fixed approach for any EIS plugged into any Application Server, while it allowed custom interface defined between Application and Resource Adaptor, i.e. EIS.

Reference: "This document begins by describing the rationale and goals for creating a standard architecture to integrate an application server with multiple heterogeneous EISs."

An application server and an EIS collaborate to keep all system-level mechanisms—transactions, security, and connection management—transparent from the application components.

Connector Architecture------An architecture for integrating J2EE servers with EISs. There are two parts to this architecture: an EIS vendor-provided resource adapter and an application server that allows this resource adapter to be plugged in. This architecture defines a set of contracts (such as transactions, security, connection management) that a resource adapter has to support to plug in to an application server. These contracts support bi-directional communication (outbound and inbound) between an application server and an EIS via a resource adapter. That is, the application server may use the resource adapter for outbound communication to the EIS, and it may also use the resource adapter for inbound communication from the EIS.

The "Inbound/Outbound" is refer to direction between application and resource adapter, not resource adapter and EIS.

In fact, almost all contracts are between application server and resource adapter, or between application and resource adapter, the contracts between resource adapter and EIS is private, or vendor-specified.

The resource adapter lists the types of message listener that it supports. The J2EE Connector Architecture version 1.5 and the EJB version 2.1 specifications do not restrict message listeners to using the JMS API.

That means, MDB need not be related with JMS, in fact, MDB is related with Resource Adaptor.

2. Contracts

Contract	The two parties	Goals
Lifecycle management contract	application server and resource adapter	allows an application server to manage the lifecycle of a resource adapter
Work management contract	application server and resource adapter	allows a resource adapter to do work (monitor network endpoints, call application components, etc.) by submitting Work instances to an application server for execution
Transaction inflow contract	application server and resource adapter	allows a resource adapter to propagate an imported transaction to an application server.
Message inflow contract	application server and resource adapter	allows a resource adapter to asynchronously deliver messages to message endpoints residing in the application server. This contract also serves as the standard message provider pluggability contract that allows a wide range of message providers to be plugged into any J2EE compatible application server via a resource adapter

3. CCI

Rather than using CCI directly, developers will use EIS specific tools which based on CCI to access EIS, such as JDBC API, JMS API

The JDBC interfaces—javax.sql.DataSource, java.sql.Connection—are examples of non-CCI connection factory and connection interfaces. Note that the methods defined on a non-CCI interface are not required to throw a ResourceException. The exception can be specific to a resource adapter, for example: java.sql.SQLException for JDBC interfaces.

Client API------The client API used by application components for EIS access may be defined as: The standard Common Client Interface (CCI)

A client API specific to the type of a resource adapter and its underlying EIS. An example of such an EIS specific client API is JDBC for relational databases. The Common Client Interface (CCI) defines a common client API for accessing EISs.
The CCI is targeted towards Enterprise Application Integration (EAI) and enterprise tools vendors.

4. Resource Adapter

The ResourceAdapter JavaBean represents a resource adapter instance and contains the configuration information pertaining to that resource adapter instance. This configuration information may also be used as global defaults for ManagedConnectionFactory and ActivationSpec JavaBeans. That is, when ManagedConnectionFactory or ActivationSpec JavaBeans are created they may inherit the global defaults (ResourceAdapter JavaBean configuration information), which make it easier to configure them. A resource adapter instance may provide bi-directional connectivity to multiple EIS instances. A ManagedConnectionFactory JavaBean can be used to provide outbound connectivity to a single EIS instance. An ActivationSpec JavaBean can be used to provide inbound connectivity from an EIS instance. A resource adapter instance may contain several such ManagedConnectionFactory and ActivationSpec JavaBeans.

from perspective of Application:

ManagedConnectionFactory : 送站通道

ActivationSpec : 接站通道

ResourceAdaptor JavaBean is not full-functional Resource Adaptor, it must combine with ManagedConnectionFactory JavaBean, ActivationSpec JavaBean and other JavaBeans to serve as a full-functional Resource Adaptor

The CCI defines javax.resource.cci.ConnectionFactory and javax.resource.cci.Connection as interfaces for a connection factory and a connection, respectively.

The connection factory instance delegates the connection creation request to the ConnectionManager instance.

The ConnectionManager instance, on receiving a connection creation request from the connection factory, does a lookup in the connection pool provided by the application server. If there is no connection in the pool that can satisfy the
connection request, the application server uses the ManagedConnectionFactory interface (implemented by the resource adapter) to create a new physical connection to the underlying EIS.

The application server registers a ConnectionEventListener with the ManagedConnection instance. This listener enables the application server to get event notifications related to the state of the ManagedConnection instance.

The application server uses the ManagedConnection instance to get a connection instance that acts as an application-level handle to the underlying physical connection. An instance of type javax.resource.cci.Connection is an example
of such a connection handle. An application component uses the connection handle to access EIS resources.

The circuitous route is for reusing application server provided services, such as transaction management, security, error logging and tracing, and connection pool management.

In the JDBC architecture, an application server provides the implementation of javax.sql.DataSource interface. Using a similar design approach for the connector architecture would have required an application server to provide implementations of various connection factory interfaces defined by different resource adapters. Since the connection factory interface may be defined as specific to an EIS, the application server may find it difficult to provide implementations of connection factory interfaces without any code generation.

The connection factory implementation class delegates the getConnection method invocation from an application component to the associated ConnectionManager instance. The ConnectionManager instance is associated with a connection factory
instance at its instantiation

Note that the connection factory implementation class must call the ConnectionManager.allocateConnection method in the same thread context in which the application component had called the getConnection method.

A resource adapter must provide a default implementation of the javax.resource.spi.ConnectionManager interface. The implementation class comes into play when a resource adapter is used in a non-managed two-tier application scenario. In an application server-managed environment, the resource adapter must not use the default ConnectionManager implementation class.

5. Transaction

The ERP system supports JTA transactions by implementing a XAResource interface through its resource adapter. The TP system also implements an XAResource interface. This interface enables the two resource managers to participate in
transactions that are coordinated by an external transaction manager. The transaction manager uses the XAResource interface to manage transactions across the two underlying resource managers.

If multiple resource managers participate in a transaction, the EJB container uses a transaction manager to coordinate the transaction. The contract between the transaction manager and resource manager is defined using the XAResource
interface. If a single resource manager instance participates in a transaction (either componentmanaged or container-managed), the container has two choices:

It uses the transaction manager to manage this transaction. The transaction manager uses one-phase commit-optimization to coordinate the transaction for this single resource manager instance.
The container lets the resource manager coordinate this transaction internally without involving an external transaction manager.

The Java Transaction API (JTA) is a specification of interfaces between a transaction manager and the other parties involved in a distributed transaction processing system: application programs, resource managers, and an application
server.

The Java Transaction Service (JTS) API is a Java binding of the Common Object Request Broker Architecture (CORBA) Object Transaction Service (OTS) 1.1 specification. JTS provides transaction interoperability using the standard Internet
Inter-ORB Protocol (IIOP) for transaction propagation between servers. The JTS API is intended for vendors who implement transaction processing infrastructure for enterprise middleware. For example, an application server vendor can use a JTS
implementation as the underlying transaction manager.

The interface ManagedConnection defines method associateConnection as follows:

public interface javax.resource.spi.ManagedConnection {
public void associateConnection(Object connection) throws ResourceException;
...
}

The container uses the associateConnection method to change the association of an application-level connection handle with a ManagedConnection instance. The container finds the right ManagedConnection instance, depending on the connection sharing scope, and calls the associateConnection method. To achieve this, the container needs to keep track of connection handles acquired by component instances and ManagedConnection instances using an implementation-specific mechanism.

The associateConnection method implementation for a ManagedConnection should dissociate the connection handle passed as a parameter from its currently associated ManagedConnection and associate the new connection handle with itself.

Note that the switching of connection associations must happen only for connection handles and ManagedConnection instances that correspond to the same ManagedConnectionFactory instance. The container should enforce this restriction
in an implementation-specific manner. If a container cannot enforce the restriction, the container should not use the connection association mechanism.

The resource adapter must implement the associateConnection method to support connection sharing. The container makes a decision on whether or not to use the associateConnection method implemented by a resource adapter.

6. Security

If the res-auth element is set to Application, the component code performs a programmatic sign-on to the EIS.

If the res-auth element is Container, the application server takes on the responsibility of setting up and managing EIS sign-on.

This security contract supports EIS sign-on by:

Passing the connection request from the resource adapter to the application server, enabling the application server to hook-in security services.
Propagation of the security context, that is, JAAS Subject with principal and credentials, from the application server to the resource adapter.

The application server has the following options for invoking the createManagedConnection method:

Option A: The application server invokes the createManagedConnection method by passing in a non-null Subject instance that carries a single resource principal and its corresponding password-based credentials, represented by the class PasswordCredential that provides the user name and password. The PasswordCredential should be set in the Subject instance as part of the private credential set. Note that the passed Subject can contain multiple PasswordCredential instances.
The resource adapter extracts the username and password from this Subject instance by looking for the PasswordCredential instance in the Subject, and uses this security information to sign-on to the EIS instance during connection creation.

Option B: The application server invokes the createManagedConnection method by passing in a non-null Subject instance that carries a single resource principal and its security credentials. In this option, credentials are represented through the GSSCredential interface. A typical example is a Subject instance with Kerberos credentials. For example, an application server may use this option for createManagedConnection method invocation when the resource principal is
impersonating the caller or initiating principal, and has valid credentials acquired through impersonation. An application server may also use this option for principal mapping scenarios with credentials of a resource principal represented through the GSSCredential interface. Note that sensitive credentials requiring special protection, such as private cryptographic keys, are stored within a private credential set, while credentials intended to be shared, such as public key certificates or Kerberos server tickets, are stored within a public credential set. The two methods
getPrivateCredentials and getPublicCredentials should be used accordingly. In the case of Kerberos mechanism type, the application server must pass the principal’s ticket granting ticket (TGT) to a resource adapter in a private credential set. The resource adapter uses the resource principal and its credentials from the Subject instance to go through the EIS sign-on process before creating a new connection to the EIS.

Option C: The application server invokes the createManagedConnection method by passing a null Subject instance. The application server must use this option for the component-managed sign-on case. In this option, security information is carried in the ConnectionRequestInfo instance. The application server does not provide any security information that can be used by the resource adapter for managing EIS sign-on. During the deployment of a resource adapter, the application server must be configured to use one of the above specified invocation options.

A resource adapter can do EIS sign-on and connection creation in an implementation-specific way, or it can use the GSS-API. A resource adapter has the following options, corresponding to the options for an application server, for handling the invocation of the createManagedConnection method:

Option A: The resource adapter explicitly checks whether the passed Subject instance carries a PasswordCredential instance using the Subject.getPrivateCredentials method. Note that the security contract assumes that a resource adapter has the necessary security permissions to extract a private credential set from a Subject instance. The specific mechanism through which such permission is set up is outside the scope of the connector architecture. If the Subject instance contains a PasswordCredential instance, the resource adapter extracts the username and password from the PasswordCredential. It uses the security information to authenticate the resource principal, corresponding to the username, to the EIS during the creation of a connection. In this case, the resource adapter uses an authentication mechanism that is EIS specific. Since a Subject instance can carry multiple PasswordCredential instances, a ManagedConnectionFactory should only use a PasswordCredential instance that has been specifically passed to it through the security contract. The getManagedConnectionFactory method enables a ManagedConnectionFactory instance to determine whether or not a PasswordCredential instance is to be used for sign-on to the target EIS instance. The ManagedConnectionFactory implementation uses the equals method to compare itself with the passed instance.

Option B: The resource adapter explicitly checks whether the passed Subject instance carries a GSSCredential instance using the getPrivateCredentials and getPublicCredentials methods defined in the Subject interface. In the case of Kerberos mechanism type, the resource adapter must extract Kerberos credentials using the getPrivateCredentials method in the Subject interface. The resource adapter uses the resource principal and its credentials, represented by the GSSCredential interface, in the Subject instance to go through the EIS sign-on process. For example, this option is used for Kerberos-based credentials that have been acquired by the resource principal through impersonation. A resource adapter uses the getter methods defined in the GSSCredential interface to extract information about the credential and its principal. If a resource adapter is using the GSS mechanism, the resource adapter uses a reference to the GSSCredential instance in an opaque manner and is not required to understand any mechanism-specific credential representation. However, a resource adapter may need to interpret credential representation if the resource adapter initiates authentication in an implementation-specific manner.

Option C: If the application server invokes the ManagedConnectionFactory.createManagedConnection method with a null Subject instance, a resource adapter has the following options: The resource adapter should extract security information passed through the ConnectionRequestInfo instance. The resource adapter should authenticate the resource principal by combining the configured security information on the ManagedConnectionFactory instance with the security information passed through the ConnectionRequestInfo instance. The default behavior for the resource adapter is to allow the security information in the ConnectionRequestInfo parameter to override the configured security information in the ManagedConnectionFactory instance. If the resource adapter does not find any security configuration in the ConnectionRequestInfo instance, the resource adapter uses the default security configuration in the ManagedConnectionFactory instance. If the EIS does not require authentication, the resource adapter does not need any security information from the ConnectionRequestInfo instance, and hence may ignore such security information. This may happen due to a disconnect between the application and the resource adapter.

7. Activation

An ActivationSpec JavaBean instance encapsulates the configuration information needed to setup asynchronous message delivery to a message endpoint.

Each ActivationSpec can also specify a list of required properties. These required properties can be used to validate the configuration of an ActivationSpec JavaBean instance.

The message-driven bean developer provides activation configuration information in the message-driven bean deployment descriptor. This includes messaging style specific configuration details, and possibly message provider-specific details as well, which is used by the message-driven bean deployer to setup the activation.

The resource adapter provides the following information via the resource adapter deployment descriptor that is used by the endpoint deployer to setup endpoint activation:

List of Supported Message Listener Types
ActivationSpec JavaBean
Administered Objects

The endpoint application’s activation-config properties, specified in the endpoint deployment descriptor, should be a subset of the ActivationSpec JavaBean’s properties. There must be a one-to-one correspondence between the activation-config property names and the ActivationSpec JavaBean’s property names. This allows automatic merging of the activation-config properties with an ActivationSpec JavaBean instance during endpoint deployment. Any specified activation-config property which does not have a matching property in the ActivationSpec JavaBean should be treated as an error.

Activation Spec Value Priority	Activation Spec Value Position
Highest	MDB Configured Deployment Descriptor
	MDB Default Deployment Descriptor(Merged ActivationSpec Bean)
	Configured ActivationSpec Bean
Lowest	Default ActivationSpec Bean

Before any messages can be delivered to a message endpoint, the message endpoint must be associated with a destination. This task is performed during application installation. Therefore, the responsibility of associating a message-driven bean with a destination lies with the application deployer. The application deployer creates an instance of the ActivationSpec JavaBean for the relevant resource adapter and associates it with the message endpoint during installation. In this way an ActivationSpec JavaBean, through its destination property, associates a message endpoint with a destination on the message provider.

A message endpoint is activated by the application server when the message endpoint application is started. During message endpoint activation, the application server passes the ActivationSpec JavaBean, and a reference to the
MessageEndpointFactory, to the resource adapter by invoking its endpointActivation method.

The resource adapter uses the information in the ActivationSpec JavaBean to interact with messaging provider and setup message delivery to the message endpoint. For a JMS message-driven bean, this might involve configuring a
message selector or a durable subscription against the destination. Once the endpointActivation method returns, the message endpoint is ready to receive messages.

The following steps describe the sequence of events that occur when a message arrives at a destination:

The resource adapter detects the arrival of a message at the destination.
The resource adapter invokes the createEndpoint method on the MessageEndpointFactory.
The MessageEndpointFactory obtains a reference to a message endpoint. This might be an unused message endpoint obtained from a pool or, if no message endpoints are available, it can create a new message endpoint.
The MessageEndpointFactory returns a proxy to this message endpoint instance to the resource adapter.
The resource adapter uses the message endpoint proxy to deliver the message to the message endpoint.

Several other messaging providers exist that require similar functionality to message-driven beans within the EJB container, such as the Java API for XML Messaging (JAXM). Because of this, version 2.1 of the EJB specification expanded the definition of message-driven beans to provide support for messaging providers other than JMS providers. It does this by allowing a message-driven bean to implement an interface other than the javax.jms.MessageListener interface. The type of message listener interface that a message-driven bean implements determines its type.

The EJB specification states that, if the message listener interface supports the request-reply pattern in this manner, it is the responsibility of the EJB container to deliver the reply message to the resource adapter.

8. Work

Note that, rather than calling the MessageEndpointFactory directly, the resource adapter is likely to instantiate a Work object, and pass it to the application server via the WorkManager interface. When the doWork method of the WorkManager is called the dispatch will occur. This will allow the resource adapter to continue to process incoming messages without blocking until message-driven bean dispatch has completed.

The resource adapter can use the message inflow contract to call a message-driven bean, and use the message-driven bean to dispatch calls to session and entity beans using the EJB client view model. The EJB specification allows a request-response style message-driven bean call which could be used for synchronous RPC-style calls. The message-driven bean could be packaged either with the resource adapter or separately.

9. CCI

The CCI is designed with the following goals:

It is targeted primarily towards application development tools and EAI frameworks, not application component.
It defines a remote function-call interface that focuses on executing functions on an EIS and retrieving the results. The CCI can form a base level API for EIS access on which higher level functionality can be built.
Although it is simple, it has sufficient functionality and an extensible application programming model.
It provides an API that both leverages and is consistent with various facilities defined by the J2SE and J2EE platforms.
It is independent of a specific EIS. For example, it does not use data types specific to an EIS. However, the CCI can be capable of being driven by EIS-specific metadata from a repository.

An important goal for the CCI is to complement existing standard JDBC API and not to replace this API. The CCI defines a common client API that is parallel to the JDBC for EISs that are not relational databases.

An EAI or application development tool uses a metadata repository to drive CCIbased interactions with heterogeneous EISs. A repository may maintain meta information about functions, with type mapping information and data structures for the invocation parameters, existing on an EIS system.

The application development tool generates Java classes based on the meta information accessed from a metadata repository. These Java classes encapsulate CCI-based interactions and expose a simple application programming model, typically based on the JavaBeans framework, to the application developers. An application component uses the generated Java classes for EIS access.

The application programming model for a Record is as follows:

A component creates an instance of a generated implementation class for a custom record. The implementation class represents an EIS-specific data structure.
A component uses the RecordFactory interface to create an instance of the generic Record implementation class. The implementation class of a generic Record is independent of any EIS-specific data structure.

In cases of both custom and generic Records, the type mapping information is provided by a metadata repository either at development-time or runtime.

It is important to note that the properties passed through the getConnection method should be client-specific, such as username, password, and language, and not be related to the configuration of a target EIS instance, such as port number or server name. The ManagedConnectionFactory instance is configured with a complete set of properties required for the creation of a connection to an EIS instance. Configured properties on a ManagedConnectionFactory can be overridden by client-specific properties passed by an application component through the getConnection method.

An important point to note is about the relationship between ConnectionSpec and ConnectionRequestInfo. The ConnectionSpec is used at the application level and is defined under the scope of CCI while ConnectionRequestInfo is defined as part of the system contracts. Separate interfaces have been defined to ensure the separation between CCI interfaces and system contracts. ConnectionRequestInfo has no explicit dependency on CCI. Note that a resource adapter may not implement CCI but it must implement system contracts. The specification of a standard repository API and metadata format is outside the scope of the current version of the connector architecture. The mapping between CCI’s ConnectionSpec and ConnectionRequestInfo is achieved in an implementation-specific manner by a resource adapter.

10. Deployment

When a standalone resource adapter RAR is deployed, the resource adapter must be made available to all J2EE applications in the application server. When a resource adapter RAR packaged within a J2EE application EAR is deployed, the resource adapter must be made available only to the J2EE application with which it is packaged.

The deployment code retrieves the configuration properties from the XML deployment descriptor for the resource adapter. The deployment code then creates an instance of the ManagedConnectionFactory implementation class and configures the properties of the instance.

The deployment code uses the ManagedConnectionFactory instance to create a connection factory instance. The code then registers the connection factory instance in the JNDI namespace.

The following steps occur when an application component calls the method JNDI Context.lookup to lookup a connection factory instance:

JNDI passes control to the application server. The ObjectFactory.getObjectInstance method implemented by the application server is called.
The application server creates a new instance of the ManagedConnectionFactory implementation class provided by the resource adapter.
The application server calls setter methods on the ManagedConnectionFactory instance to set various configuration properties of this instance. These properties provide information required by the ManagedConnectionFactory instance to
create physical connections to the underlying EIS. The application server uses an existing property set configured during the deployment of a resource adapter to set the required properties of the ManagedConnectionFactory instance.
After the newly created ManagedConnectionFactory instance has been configured with its properties set, the application server creates a new ConnectionManager instance.
The application server calls the createConnectionFactory method of the ManagedConnectionFactory instance, passing in the ConnectionManager instance from the previous step, to get a ConnectionFactory instance.
The application server returns the connection factory instance to the JNDI provider, so that this instance can be returned as a result of the JNDI lookup. The application component gets the ConnectionFactory instance as a result of the
JNDI lookup.

ra.xml specify ManagedConnectionFactory-->Application Server Reflect-->associate ConnectionManager-->invoke createConnectionFactory to get cci.ConnectionFactory cccf-->client invoke cccf.getConnection-->delegate to ConnectionManager.allocateConnection-->hook service-->delegate to ManagedConnectionFactory.createConnection

切尔斯基 2006-01-12 11:22 发表评论

切尔斯基 — Wed, 28 Dec 2005 14:07:00 GMT

摘要: 1. WebSphere with Optimizeit
2. WebSphere with DataSource
3. WebSphere with JVM Arguments
4. WebSphere with Default JMS Provider, JCA, and SIB
5. WebSphere with MQSeries
6. WebSphere with JNDI
7. Misc 阅读全文

切尔斯基 2005-12-28 22:07 发表评论

亚洲丰满熟女一区二区哦,亚洲18在线天美,亚洲成人高清在线

Restatement: 性能,定w��,负蝲,以及(qi��ng)压力���试

1, 部��v时如何方便的支持产品环境和测试环�? 卛_��不同环境下提供同一接口的不同实�?/h3> �q�个甚至不需要IoC框架解决, 应用开发�h员在不同环境下提供不同配�|�文件即�? 那问题变成了(ji��n)

2, 如何���化配�|? 使得框架能够自动识别依赖

4, 多用��L(f��ng)��境下, 如何保证每个用户都有自己的对象组合而不互相�q�扰

如何不带来额外的效率损失

如何方便查找特定条�g的错�?/h3> 查找是数据库的强��? 把日志结构化, 存到数据库里卛_��. �q�会(x��)带来事务的问�? 写日志是否和用户正常的业务放在一个事务里?

如何横切的添加通用信息

谁来做Validation

何时做Validation

如何表达错误

如何传递错�?/h3> 攉���参数, 输出参数, Thread Local, 或者抛出异常然后合适的层次捕获

如何兌���错误到发生错误的对象�? ���其是对象图中非Root对象

一个额外的话题

Web开发问题域

跨团队的持箋(hu��)集成: 几个基本矛盾

ThrowawayObject :在Java中实现Rails里的Flash Message

敏捷质疑: �l�对�~�程, 代码集体所有权

敏捷质疑: 持箋(hu��)集成

敏捷质疑: TDD

CruiseControl Enterprise 最�?j��ng)_���?(6) : Scaling up

CruiseControl Enterprise 最�?j��ng)_���?(5) : Refactor your configuration file

CruiseControl Enterprise 最�?j��ng)_���?(4) : Bootstrap with a Bootstrapper

Ivy: 搭徏本地仓库

仓库本��n

如何使用�q�个仓库

Ivy 如何解决我们的问�?/h3> Q: 我的��目在开发环境和产品环境有不同的依赖, 怎么�?

RegexBuilder: A DSL for creating regular expression, Request For Comment

CruiseControl Enterprise 最�?j��ng)_���?(3) : Configuring CruiseControl the CruiseControl way

CruiseControl Enterprise 最�?j��ng)_���?(2) : Keep your dependencies to yourself

CruiseControl Enterprise 最�?j��ng)_���?(1) : Publish with a Publisher

jadestone enhanced sax handlers 2.0 发布

[导入]���单至�?qi��ng)的AOP和IOC

[导入]Configuration Pattern : 配置模式

[导入]Axis : �q�去前的备忘

Enhanced SAX Handler :�?DOM �q�简单的 SAX Handler

emma : ���试覆盖率低于指标的时候让Ant Build自动��p�|

RegexBuilder: A DSL for creating regular expression

Get Mac Address in Java

WS-Security Practice

Essential Vitria BusinessWare

Essential Web Services: SOAP, WSDL, UDDI

I. Questions before reading the specification, and Answers after reading the specification.

II. Notes

Restatement: 性能,定w��,负蝲,以及(qi��ng)压力��试

1, 部��v时如何方便的支持产品环境和测试环�? 卛_��不同环境下提供同一接口的不同实�?/h3>
�q�个甚至不需要IoC框架解决, 应用开发�h员在不同环境下提供不同配�|�文件即�? 那问题变成了(ji��n)

2, 如何��化配�|? 使得框架能够自动识别依赖

如何方便查找特定条�g的错�?/h3>
查找是数据库的强��? 把日志结构化, 存到数据库里卛_��. �q�会(x��)带来事务的问�? 写日志是否和用户正常的业务放在一个事务里?

如何传递错�?/h3>

攉��参数, 输出参数, Thread Local, 或者抛出异常然后合适的层次捕获

如何兌��错误到发生错误的对象�? ��其是对象图中非Root对象

CruiseControl Enterprise 最�?j��ng)_��?(6) : Scaling up

CruiseControl Enterprise 最�?j��ng)_��?(5) : Refactor your configuration file

CruiseControl Enterprise 最�?j��ng)_��?(4) : Bootstrap with a Bootstrapper

CruiseControl Enterprise 最�?j��ng)_��?(3) : Configuring CruiseControl the CruiseControl way

CruiseControl Enterprise 最�?j��ng)_��?(2) : Keep your dependencies to yourself

CruiseControl Enterprise 最�?j��ng)_��?(1) : Publish with a Publisher

[导入]��单至�?qi��ng)的AOP和IOC

emma : ��试覆盖率低于指标的时候让Ant Build自动��p�|