写一个简单的工作流（二）

hoho，今天完成了选择路由的实现，完成了配置文件的读写和解析，流程定义文件还是决定采用xml文件，不过与其他工作流引擎采用的xml完全不同，因为是基于petri网的，因此引入了place的概念，比如下面这个4个节点的顺序路由的流程： < workflow  maxCases ="100" >      < node  type ="start"  name ="start"  id ="0" >          < inputs >              < place  id ="1"   />          </ inputs >          < outputs >              < place  id ="2"   />          </ outputs >      </ node >      < node  name ="hello"  id ="1"  resource ="user" >          < conditions  type ="and" >              < condition                  class ="net.rubyeye.insect.workflow.impl.NullHandler"  value ="false"                 variable-name ="name"   />          </ conditions >          < handler              class ="net.rubyeye.insect.workflow.test.HelloWorldHandler"   />          < inputs >              < place  id ="2"   />          </ inputs >          < outputs >              < place  id ="3"   />          </ outputs >      </ node >      < node  name ="calc"  id ="2"  resource ="user" >          < conditions  type ="and" >              < condition  variable-name ="num" >                  < exp >                      <![CDATA[ num<=1000 ]]>                  </ exp >              </ condition >              < conditions  type ="or" >                  < condition  variable-name ="num" >                      < exp >                          <![CDATA[ num>=10 ]]>                      </ exp >                  </ condition >                  < condition                      class ="net.rubyeye.insect.workflow.impl.NullHandler"  value ="false"                     variable-name ="name"   />              </ conditions >          </ conditions >          < handler              class ="net.rubyeye.insect.workflow.test.CalculateHandler"   />          < inputs >              < place  id ="3"   />          </ inputs >          < outputs >              < place  id ="4"   />          </ outputs >      </ node >      < node  type ="end"  name ="hello"  id ="3" >          < inputs >              < place  id ="4"   />          </ inputs >          < outputs >              < place  id ="5"   />          </ outputs >      </ node > </ workflow >     并行路由和选择路由引入了and-split,and-join,or-split,or-join四种transition，比如and-split，它就有多个输出place： < node  name ="split"  type ="and-split"  id ="1"  resource ="user" >          < inputs >              < place  id ="2"   />          </ inputs >          < outputs >              < place  id ="3"   />              < place  id ="4"   />          </ outputs >      </ node >    Place和Transition都有条件，用于决定操作是否执行，Transition额外指定了驱动的资源以及回调的handler，这一点非常重要，资源可能是用户、用户组、某个时间点定时事件、特定消息等等。    今天额外发现的一个好处就是，引入Place之后，我可以轻易地将不同的流程连接起来组织成一个更复杂的流程，仅仅是需要修改各个流程的开始和结束的节点的输入输出库所,从而实现了层次化的Petri网,，类似代码： WorkFlow sequence  =  wm.getWorkFlow( " sequence " );         WorkFlow concurrency  =  wm.getWorkFlow( " concurrency " );         WorkFlow choose  =  wm.getWorkFlow( " choose " );          // 组合流程         composite  =   new  WorkFlow();         composite.setName( " composite " );         composite.setId( 100 );                  wm.saveWorkFlow(composite);          // 修改开始结束节点的输入输出库所         sequence.getEnd().setType(TransitionType.NORMAL);         sequence.getEnd().setOutputs(concurrency.getStart().getInputs());         concurrency.getEnd().setType(TransitionType.NORMAL);         concurrency.getEnd().setOutputs(choose.getStart().getInputs());                  composite.setStart(sequence.getStart());         composite.setEnd(choose.getEnd());         List < Transition >  transitions  =   new  ArrayList < Transition > ();         transitions.addAll(sequence.getTransitions());         transitions.addAll(concurrency.getTransitions());         transitions.addAll(choose.getTransitions());         composite.setTransitions(transitions);

文章转自庄周梦蝶  ，原文发布时间2007-10-11