activiti源码解读之心得整编-阿里云开发者社区

activiti源码解读之心得整编

2017-09-16 1629

版权

本文内容由阿里云实名注册用户自发贡献，版权归原作者所有，阿里云开发者社区不拥有其著作权，亦不承担相应法律责任。具体规则请查看《阿里云开发者社区用户服务协议》和《阿里云开发者社区知识产权保护指引》。如果您发现本社区中有涉嫌抄袭的内容，填写侵权投诉表单进行举报，一经查实，本社区将立刻删除涉嫌侵权内容。

简介：

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：
[java] view plain copy print ?
1. // Push on stack
2. Context.setCommandContext(context);
3. Context.setProcessEngineConfiguration(processEngineConfiguration);
5. return next.execute(config, command);
这边push，另外有地方pop，CommandInvoker就干的此事：
[java] view plain copy print ?
1. public T execute(CommandConfig config, Command command) {
2. return command.execute(Context.getCommandContext());
3. }
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：
[java] view plain copy print ?
1. protectedvoid eventNotificationsCompleted(InterpretableExecution execution) {
2. TransitionImpl transition = execution.getTransition();
3. ActivityImpl destination = null;
4. if(transition == null) { // this is null after async cont. -> transition is not stored in execution
5. destination = (ActivityImpl) execution.getActivity();
6. } else {
7. destination = transition.getDestination();
8. }
9. ActivityImpl activity = (ActivityImpl) execution.getActivity();
10. if (activity!=destination) {
11. ActivityImpl nextScope = AtomicOperationTransitionNotifyListenerTake.findNextScope(activity, destination);
12. execution.setActivity(nextScope);
13. execution.performOperation(TRANSITION_CREATE_SCOPE);
14. } else {
15. execution.setTransition(null);
16. execution.setActivity(destination);
17. execution.performOperation(ACTIVITY_EXECUTE);
18. }
19. }
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：
[java] view plain copy print ?
1. List joinedExecutions = executionEntity.findInactiveConcurrentExecutions(execution.getActivity());
2. if (joinedExecutions.size() == nrOfInstances || completionConditionSatisfied(execution)) {
4. // Removing all active child executions (ie because completionCondition is true)
5. List executionsToRemove = new ArrayList ();
6. for (ActivityExecution childExecution : executionEntity.getParent().getExecutions()) {
7. if (childExecution.isActive()) {
8. executionsToRemove.add((ExecutionEntity) childExecution);
9. }
10. }
11. for (ExecutionEntity executionToRemove : executionsToRemove) {
12. if (LOGGER.isDebugEnabled()) {
13. LOGGER.debug("Execution {} still active, but multi-instance is completed. Removing this execution.", executionToRemove);
14. }
15. executionToRemove.inactivate();
16. executionToRemove.deleteCascade("multi-instance completed");
17. }
18. 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：
[java] view plain copy print ?
1. protectedvoid ensureProcessDefinitionInitialized() {
2. if ((processDefinition == null) && (processDefinitionId != null)) {
3. ProcessDefinitionEntity deployedProcessDefinition = Context
4. .getProcessEngineConfiguration()
5. .getDeploymentManager()
6. .findDeployedProcessDefinitionById(processDefinitionId);
7. setProcessDefinition(deployedProcessDefinition);
8. }
9. }
11. protectedvoid ensureActivityInitialized() {
12. if ((activity == null) && (activityId != null)) {
13. activity = getProcessDefinition().findActivity(activityId);
14. }
15. }
再来看看execution的set方法，就能明白它为什么会保留一堆id：
[java] view plain copy print ?
1. publicvoid setActivity(ActivityImpl activity) {
2. this.activity = activity;
3. if (activity != null) {
4. this.activityId = activity.getId();
5. this.activityName = (String) activity.getProperty("name");
6. } else {
7. this.activityId = null;
8. this.activityName = null;
9. }
10. }
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：
[java] view plain copy print ?
1. publicvoid execute(InterpretableExecution execution) {
2. InterpretableExecution propagatingExecution = null;
3. ActivityImpl activity = (ActivityImpl) execution.getActivity();
4. if (activity.isScope()) {
5. propagatingExecution = (InterpretableExecution) execution.createExecution();
6. propagatingExecution.setActivity(activity);
7. propagatingExecution.setTransition(execution.getTransition());
8. execution.setTransition(null);
9. execution.setActivity(null);
10. execution.setActive(false);
11. log.debug("create scope: parent {} continues as execution {}", execution, propagatingExecution);
12. propagatingExecution.initialize();
14. } else {
15. propagatingExecution = execution;
16. }
18. propagatingExecution.performOperation(AtomicOperation.TRANSITION_NOTIFY_LISTENER_START);
19. }
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：
[java] view plain copy print ?
1. publicclass ActivityInstanceEndHandler implements ExecutionListener {
3. publicvoid notify(DelegateExecution execution) {
4. Context.getCommandContext().getHistoryManager()
5. .recordActivityEnd((ExecutionEntity) execution);
6. }
7. }