Integration for jmolecules-axon

[zambrovski]

Hi folks,

you are probably aware of Axon Framework.
I would like to suggest to create a jmolecules-axon-integration.

Is it something you are interested in?

Cheers,

Simon

[hschwentner]

Hi Simon,

I think @abuijze would agree, right? :-)

Best,
Henning

[zambrovski]

I don't think he will prohibit to to do so ;-)

[zambrovski]

Hi Henning,

I would like to have a closer discussion on that - I had a review on DDD, CQRS and EVENT annotations/types, but still miss some...
Is there any further documentation on that available?

I understand Command and CommandHandler and QueryModel.

I don't yet understand the CommandDispatcher... Is it a place responsible for sending commands (usually then something inside the application ring or domain service ring) or is it a place making sure the commands are dispatched to the correct handlers?

I also believe it would be nice to have a CommandModel as an opposite of the same granularity to the QueryModel. The reason for that is, that a command handler is a method/function/operation as a part of aggregate/domain entity/domain service... But I'd like to mark the components responsible for the command validation and processing ... Aggregate might be too specific for this... I usually have a simple annotation CommandModel for marking the part of the CQRS domain core, representing the "write model".

Additional stuff I would like to propose is QueryHandler (logical symmetry to command handler and event handler) as a marker for the method/function/operation responsible for answering the queries... The next step would be to introduce Query to allow materialized query objects to be identified... From these two proposals, QueryHandler is already implemented in Axon Framework, and we implemented our own Query marker in our Axon projects ...

What do you think?

Cheers,

Simon

[OLibutzki]

I don't yet understand the CommandDispatcher... Is it a place responsible for sending commands (usually then something inside the application ring or domain service ring) or is it a place making sure the commands are dispatched to the correct handlers?

It's not trivial to identify which component emits a query, command or event, because usually you do not have an explicit annotation of the publishing side.

We decided to handle the construction of a command, query or event as publishing. Why shall you instaniate an event if you don't want to publish it? This heuristic approach works quite well for us.

[zambrovski]

Ok, so the source for Commands, Events and Queries will be marked with x-Dispatcher... Sounds cool...

[OLibutzki]

We do not use the CommandDispatcher or DomainEventPublisher at all. There is no need for this explicit information as we derive it from the command's or event's construction.

[OLibutzki]

@odrotbohm told me about a discussion with @abuijze to think about doing it the other way around... The axon annotations are meta-annotated using jmolecules native annotations.

From the perspective of archifacts I decided to provide an explicit module for supporting Axon...

[odrotbohm]

Please refer to #18 and #39 for the history of the CQRS annotations. I think we could make the Axon integration work through our ByteBuddy plugin by translating the jMolecules annotations into the Axon-native ones. It might shy away users who are not eager to start anything related to code generation, though. Thus, I wouldn't mind seeing a native Axon integration implemented on their side in the first place.

[mvonrenteln]

We would prefer a native implementation, too, as we are using Quarkus and Reflection/Bytecode Manipulation is discouraged there... A ByteBuddy based implementation would not be an option...

[mvonrenteln]

Great idea! We would be interested in such an integration, too. We are using Axon with Quarkus and are planning to introduce jMolecules and Arch Unit...

[odrotbohm]

The code generation would just be adding annotations to the classfiles that any native image solution should handle just like the original one. The current functionality available (generating JPA, JDBC and MongoDB boilerplate for aggregates) just works fine on e.g. Spring Native.

[zambrovski]

#Hacktober is running, I can imagine to get something done ;-)