Note on @CDIComponent with regard to Jakarte EE 10

[roeltje25]

I refer to issue #303 where I addressed that @RouteScoped and @UIScoped annotations were properly taken into account when using bean-discovery-mode="annotated". The decision was taken to require a @CDIComponent annotation for any bean with @RouteScoped annotation when using bean-discovery-mode="annotated". Those that use this discovery mode will then have to use double annotations for all instances where @RouteScoped is used.

Now with that context in mind I notice that in Jakarte EE 10 the default bean-discovery-mode is changed to "annotated". Which will mean that the impact of the requirement of needing an @CDIComponent annotation is enlarged to any and all that do not explicitly have set their bean-discovery-mode to "all".

I do dare to stubbornly but confidently suggest to mark the @RouteScoped and @UIScoped annotations @stereotype after all to prevent users to code tautologies ( extends Component and also mark CDIComponent, and also say @RouteScoped) and prevent any future problems when people migrate to Jakarta EE 10.

[roeltje25]

Anyone thinking about this issue?

[mshabarov]

@roeltje25 thanks for the issue. Yes, we are thinking how we can fix it on our side and avoid extra annotations.

[mshabarov]

The problem with bean discovery mode is a bit wider: without setting it to all Vaadin applications end up with views not discovered as CDI beans even if they have no scope annotations on top of it, see for instance this issue with Payara server vaadin/flow#15625 (comment)

[roeltje25]

I know. This is the exact "problem" or in fact feature we have dealt with all along. We always explicitly set the bean discovery mode to annotated in our projects. To us that makes more sense (performance and control). We annotate all our view with @RouteScoped or @UIScoped. Which in fact gives issues if they are not marked as @stereotype. In CDI with annotated mode I would consider it a bug if a bean is discovered while not explicitly having a bean defining annotation, so @route may not be the right candidate to become a bean defining annotation. But the Scopes suggest to be scopes, but they are not implemented as bean defining annotations, while as far as my knowledge goes, all other scopes (in CDI for sure, also the scopes in the projects I know) are bean defining as well..... except for the @RouteScoped and @UIScoped in Vaadin. This is the exact reason for my pushing for those annotations to be bean defining annotations.

[mcollovati]

I may be wrong but in CDI only @ApplicationScoped, @RequestScoped, @Dependent and normal scopes are bean defining annotations

https://jakarta.ee/specifications/cdi/4.0/jakarta-cdi-spec-4.0.html#bean_defining_annotations

And, if I recall correctly, we have normal route and ui scope.

Of course they behave differently from the non normal scope, and cannot be used on vaadin components

I am still not convinced of annotating the non normal scopes with stereotype, but this is just my personal opinion

[roeltje25]

Actually all annotations that have a @stereotype annotation will also be bean defining. That will not automatically imply they will be @normalscope. For @routescoped and @uiscoped a @stereotype annotation would make them bean defining while retaining the not-normalscope behavior. See my comments on the refered issue about that. Thanks sofar Op di 31 jan. 2023 14:47 schreef Marco Collovati ***@***.***>:

…

I may be wrong but in CDI only @ApplicationScoped, @RequestScoped, @dependent and normal scopes are bean defining annotations https://jakarta.ee/specifications/cdi/4.0/jakarta-cdi-spec-4.0.html#bean_defining_annotations And, if I recall correctly, we have normal route and ui scope. Of course they behave differently from the non normal scope, and cannot be used on vaadin components — Reply to this email directly, view it on GitHub <#411 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAQLW4UXJT6LMBDEWWDFO7TWVEJXNANCNFSM6AAAAAARBH65KA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[mcollovati]

I know that adding a stereotype makes an annotations a bean defining one. My point is that by spec only some kinds of scopes are naturally bean defining annotations and that specs says that a stereotype may have a scope but there is no mention about making a scope a stereotype.
But, again, that is only my point of view.
The team will decided how to handle this

[mshabarov]

I personally don't have enough expertise to decide to include AtStereotype or not.
This is not a priority at the moment for Flow team.
Let us keep this in a backlog until further investigation of this topic.