implement artifactFactory for spring using applicationContext

[jangalinski]

I implemented an implementation of ArtifactFactory for Spring utilizing a applicationContextAware engine Configuration.

[martinschimak]

👍 And good to know that you did it already before I started to do it.

[buildhive]

camunda BPM » camunda-bpm-platform #1690 UNSTABLE
Looks like there's a problem with this pull request
(what's this?)

[bconnection-zz]

Hello Jan,
that looks fine. I supplemented the following:

    public <T> T create (final Class<T> springManagedClass, final Object ...args){
        return this.context.getBean(springManagedClass, args);
    }

Then it is also possible to use parameters at constructors.

Thank you for your work.

Matthias

[jangalinski]

Hi Matthias I do not understand what you mean ... the signature defined by the ArtifactFactory interface is fixed ... how would you use your additional code?

[bconnection-zz]

Hello Jan,
with your code you can write things like:

@Autowired
SpringArtifactFactory factory;

void someMethod(){
    SomeClassType instance = factory.createArtifact(someClass);

With my extension you could additional write:

   SomeClassType instance = factory.createArtifact(someClass, param1, param2);

With the second variant you are able to utilize also constructors which require arguments.

Sorry, I did not apply to the identifiers you introduced in your code, instead I copied the example from my implementation.
Kind regards
Matthias

[jangalinski]

I do get what the code might be good for. But in this special case, we want to use a standard interface of the camunda platform, mplement it for spring so that when a listener or delegate has to be created for a task while executing a process, we can provide a fqn of the class and we will get not the standard constructor instance but a spring bean with injection. Your extension on the other hand is a wrapper for the applicationContext, I would not need the artifactFactory at all, I could just operate on the context directly. Or am I missing something?

[bconnection-zz]

Yes, I did the implementation, because I was not aware of the possibilities in Camunda. I posted my proposal, because I couldn't find any hints, how I could instantiate Spring classes.

I read the class string from the bpmn KeyForm attribute within a execution listener. Therefore it is my code to instantiate the classes. So the second form became useful for me.

The most tricky part was: how could I provide access to the application context, because it requires to implement the ApplicationContextAware Interface. Spring will provide the context by wiring it at startup. Thus the required scope for the factory is "singleton". But if the scope is "prototype" like the classes you will instantiate during execution you need the factory. The factory is singleton and knows the context and can create instances of arbitrary objects during execution time of the system.

To introduce it in the expressions, the parameterless form is sufficient.

[jangalinski]

"To introduce it in the expressions, the parameterless form is sufficient. " .... but we do agree, that (at least with the code base today), no other than the parameterless form is supported by either the artifactFactory interface or the camunda:extension for bpmn. So this might be a good idea in the future, but is not applicable for this PR, right?

[bconnection-zz]

Yes, for the use via Camunda.

But if you autowire the SpringArtifactFactory you may also use the parameter based variant.

[saig0]

Hi Jan,

thank you for your Pull Request. It looks good!

But before I merge it, can you please provide more unit tests for process execution (start a process instance which uses the SpringArtifactFactory for resolving the class of a service task or task listener - see ServiceTaskSpringDelegationTest for example). Additionally, please use xml configuration instead of annotations like the other test cases.

Greetings,
Philipp

[saig0]

@jangalinski: Please provide the requested changes. Otherwise, I have to close the pull request for now. Note that you can re-open it later again.

[jangalinski]

Sorry, been busy. Will provide tonight

[saig0]

perfect 👍 thank you.

[jangalinski]

I am having trouble with my setup that I wont be able to fix before the weekend. So unless someone wants to take over and provide the xml-context tests, this has to wait, sorry.

[saig0]

So we wait till next week. No problem :)

[saig0]

@jangalinski: Next release comming at the end of this month. You should provide the changes soon. Otherwise, this feature is not a part of the current release.

[saig0]

@jangalinski : I will close this PL for now. You can re-open it when you have time to complete the changes. If you complete it till wednesday I will merge it for current release.

[jangalinski]

Yes, good idea ... thanks for your patience.