topic Re: Execution doesn't stop at User Task, then blows up in Alfresco Archive

Execution doesn't stop at User Task, then blows up

mindcrime — Wed, 06 Mar 2013 22:26:02 GMT

Hello all, seeing some weirdness with Activiti 5.10, that I'm hoping someone can help me understand. I am kinda thinking this may be a bug, but hopefullyit's just something I'm missing about Activiti.The basic situation is this: I'm working on implementing error handling (technical errors) for serv

Re: Execution doesn't stop at User Task, then blows up

frederikherema1 — Thu, 07 Mar 2013 14:44:23 GMT

Expression-bases service-tasks only delegate to the expression, presuming the expression WILL NOT alter execution state (apart form execution variables) and will take the default outgoing flow after the expression has been called. This is a big difference with implementing an ActivityBehaviour, which is in control of the process-flow. An ActivityBehaviour is for advanced use cases, while the regular service-task expression/class mechanism is for incorporating your java-logic in the process.

Why not use a gateway after the service-task, based on an execution-vcariable that is set when the service-task encounters an exception inside it. Even better, you can use throw a BPMNError from the JavaDelegate when your logic threw an exception. This error can be caught by an error-catching event somewhere else and do the "handle exception" logic for you, using pure BPMN.

Re: Execution doesn't stop at User Task, then blows up

mindcrime — Thu, 07 Mar 2013 16:06:39 GMT

Expression-bases service-tasks only delegate to the expression, presuming the expression WILL NOT alter execution state (apart form execution variables) and will take the default outgoing flow after the expression has been called. This is a big difference with implementing an ActivityBehaviour, which is in control of the process-flow. An ActivityBehaviour is for advanced use cases, while the regular service-task expression/class mechanism is for incorporating your java-logic in the process.

OK, cool, that more or less explains that then. Thanks!

Why not use a gateway after the service-task, based on an execution-vcariable that is set when the service-task encounters an exception inside it. Even better, you can use throw a BPMNError from the JavaDelegate when your logic threw an exception. This error can be caught by an error-catching event somewhere else and do the "handle exception" logic for you, using pure BPMN.

Well, the documentation seems to argue against using BPMNError for dealing with "technical" errors, but that certainly is an option. It probably would have been my first option, except something I read in the docs seemed to be saying "don't do that". But if you guys think it's OK, I can go back and do it that way.

Re: Execution doesn't stop at User Task, then blows up

frederikherema1 — Fri, 08 Mar 2013 09:40:17 GMT

If it's a logical error that is expected in your process, than you can just model in in with a gateway after the service-task. If it's an error that can occur due to "technical issue" while executing the service-task (not one that is expected regularly, like e.g.. order not found) an error-event is a better choice.

Re: Execution doesn't stop at User Task, then blows up

mindcrime — Fri, 08 Mar 2013 20:35:17 GMT

If it's a logical error that is expected in your process, than you can just model in in with a gateway after the service-task. If it's an error that can occur due to "technical issue" while executing the service-task (not one that is expected regularly, like e.g.. order not found) an error-event is a better choice.

Sounds good. Thanks for the assistance!