activiti源码解读之心得整编

• TaskService.completeTask()的执行内幕是啥？

  activiti采取了command模式，completeTask会被包装成一个CompleteTaskCmd，一个Cmd执行的时候需要一些外围处理，如：log日志。activiti定义了一个拦截器链，链上的每个拦截器都有个next，会一直next执行下去。以CompleteTaskCmd为例，拦截器链为：

  logger拦截器-->spring事务拦截器-->CommandContext拦截器-->CommandInvoker拦截器

  其中CommandContext拦截器的工作主要是设置Context：

        // Push on stack
        Context.setCommandContext(context);
        Context.setProcessEngineConfiguration(processEngineConfiguration);
  
        return next.execute(config, command);

  这边push，另外有地方pop，CommandInvoker就干的此事：

    public <T> T execute(CommandConfig config, Command<T> command) {
      return command.execute(Context.getCommandContext());
    }

• 一个节点结束了，流程怎么知道往下走？

  答案是TaskEntity.completeTask()方法会调用execution.signal()-->activityBehavior.signal()-->activityBehavior.leave()方法，该方法最终会激活AtomicOperationTransitionNotifyListenerStart的eventNotificationsCompleted()方法，该方法会创建当前Transition的destination，代码如下：

    protected void eventNotificationsCompleted(InterpretableExecution execution) {
      TransitionImpl transition = execution.getTransition();
      ActivityImpl destination = null;
      if(transition == null) { // this is null after async cont. -> transition is not stored in execution
        destination = (ActivityImpl) execution.getActivity();
      } else {
        destination = transition.getDestination();
      }
      ActivityImpl activity = (ActivityImpl) execution.getActivity();
      if (activity!=destination) {
        ActivityImpl nextScope = AtomicOperationTransitionNotifyListenerTake.findNextScope(activity, destination);
        execution.setActivity(nextScope);
        execution.performOperation(TRANSITION_CREATE_SCOPE);
      } else {
        execution.setTransition(null);
        execution.setActivity(destination);
        execution.performOperation(ACTIVITY_EXECUTE);
      }
    }
  }

• 多实例任务怎么知道该loop已结束？

  多实例任务会启动多个任务和execution，调用execution.signal()-->activityBehavior.signal()-->activityBehavior.leave()，ParallelMultiInstanceBehavior.leave()其中包含如下代码：

      List<ActivityExecution> joinedExecutions = executionEntity.findInactiveConcurrentExecutions(execution.getActivity());
      if (joinedExecutions.size() == nrOfInstances || completionConditionSatisfied(execution)) {
  
        // Removing all active child executions (ie because completionCondition is true)
        List<ExecutionEntity> executionsToRemove = new ArrayList<ExecutionEntity>();
        for (ActivityExecution childExecution : executionEntity.getParent().getExecutions()) {
          if (childExecution.isActive()) {
            executionsToRemove.add((ExecutionEntity) childExecution);
          }
        }
        for (ExecutionEntity executionToRemove : executionsToRemove) {
          if (LOGGER.isDebugEnabled()) {
            LOGGER.debug("Execution {} still active, but multi-instance is completed. Removing this execution.", executionToRemove);
          }
          executionToRemove.inactivate();
          executionToRemove.deleteCascade("multi-instance completed");
        }
        executionEntity.takeAll(executionEntity.getActivity().getOutgoingTransitions(), joinedExecutions);
  

  completionConditionSatisfied()方法将用来判断是否该结束，takeAll()方法将结束当前子执行，并将主执行设置为active。

• 是否可以在运行时期新增/修改一个activity

  当然可以！但是记住，标注有当前activity的execution在后续执行和结束的时候会用到这个activity！如果发生程序关闭等情况，execution会尝试从ProcessDefinition里重新根据ID加载activity，如下所示：

    protected void ensureProcessDefinitionInitialized() {
      if ((processDefinition == null) && (processDefinitionId != null)) {
        ProcessDefinitionEntity deployedProcessDefinition = Context
          .getProcessEngineConfiguration()
          .getDeploymentManager()
          .findDeployedProcessDefinitionById(processDefinitionId);
        setProcessDefinition(deployedProcessDefinition);
      }
    }
  
      protected void ensureActivityInitialized() {
      if ((activity == null) && (activityId != null)) {
        activity = getProcessDefinition().findActivity(activityId);
      }
    }

  再来看看execution的set方法，就能明白它为什么会保留一堆id：

    public void setActivity(ActivityImpl activity) {
      this.activity = activity;
      if (activity != null) {
        this.activityId = activity.getId();
        this.activityName = (String) activity.getProperty("name");
      } else {
        this.activityId = null;
        this.activityName = null;
      }
    }

  所以，要完全保证程序认识被改造的activity的途径是：自定义ProcessDefinition，重写其findActivity()方法！

• 为什么bpmn文件是XML格式，但model记录里面却采用的是JSON格式，而deployment里又采用的是XML格式？

  不知道！真的不知道activiti为什么这么做！是想支持flex里面的JSON建模么？（如上结论主要是针对于activiti-modeler的实现，经仔细验证，activiti-engine对model的editorsource是没有任何限制的~~~）

• 进入多实例节点的时候，系统何时创建了新的子执行？

  答案是AtomicOperationTransitionCreateScope.execute()，代码摘录如下：

    public void execute(InterpretableExecution execution) {
      InterpretableExecution propagatingExecution = null;
      ActivityImpl activity = (ActivityImpl) execution.getActivity();
      if (activity.isScope()) {
        propagatingExecution = (InterpretableExecution) execution.createExecution();
        propagatingExecution.setActivity(activity);
        propagatingExecution.setTransition(execution.getTransition());
        execution.setTransition(null);
        execution.setActivity(null);
        execution.setActive(false);
        log.debug("create scope: parent {} continues as execution {}", execution, propagatingExecution);
        propagatingExecution.initialize();
  
      } else {
        propagatingExecution = execution;
      }
  
      propagatingExecution.performOperation(AtomicOperation.TRANSITION_NOTIFY_LISTENER_START);
    }

  其中的activity就是当前的节点。

• 什么时候保存历史记录信息？如：HistoricActivity

  魅力在于activity的executionListeners，代码如下：

  public class ActivityInstanceEndHandler implements ExecutionListener {
  
    public void notify(DelegateExecution execution) {
      Context.getCommandContext().getHistoryManager()
        .recordActivityEnd((ExecutionEntity) execution);
    }
  }