Re: [platform-ui-dev] Enforcing inheritance constraints to avoid imprope

[Markus Keller <markus_keller@xxxxxxxxxx>]

Thanks Dimitar, I can confirm that's exactly
why Eclipse bundles are not hermetically locked down.

It should be possible for clients to
do experiments that break the rules.
http://help.eclipse.org/luna/topic/org.eclipse.platform.doc.isv/reference/misc/api-usage-rules.html

API Tools' @no* tags unfortunately weakened
the established rules by stating restrictions, not permissions.
https://bugs.eclipse.org/220905

Markus



From:      
 Dimitar Georgiev <anakin.bg@xxxxxxxxx>
To:      
 "Eclipse Platform
UI component developers list." <platform-ui-dev@xxxxxxxxxxx>
Date:      
 2014-07-03 11:07
Subject:    
   Re: [platform-ui-dev]
Enforcing inheritance constraints to avoid improper API usage
Sent by:    
   platform-ui-dev-bounces@xxxxxxxxxxx

---

I think the reason is, that, even though the class is
not intended to be subclassed, it -may have- to be subclassed in some circumstances.
For example, there could be an API hole in Platform and a certain usecase
which was not identified by the platform team could only be achieved by
extending the class in question.
In such a scenario, if the class / method were marked final, the customer
would not be able to compile.

The same is the rationale for packages which are named "internal"
and are not supposed to be used, but still are OSGi-exported from the corresponding
bundles.

I'm just a user so a contributor or commiter could confirm or reject my
line of thought.

Regards, Dimitar


On 3.07.2014 06:59, Jeanderson wrote:
Hello folks, 

I'm currently working in a GSoC project to add Generics
support for JFace viewers.
While reviewing some code, I noticed the following pattern:
------------------------------------------------------------------------------------
/**
 * (...)
 *
 * @since (...)
 * @noextend This class is not intended to be subclassed
by clients.
 *
 */
public class A [extends B [implements C]] { ... }
------------------------------------------------------------------------------------

Is there any special reason to not use "final"
keyword? In addition, I'm not sure but I guess there are some methods documented
like the former example.

I know that our work is reviewed by someone to prevent
the addition of improper code (including the case that someone use bad
sub-typing). 
However, why waste time manually checking such constraints
while the compiler could save time doing this?

I don't know if this is an issue at all, I would like
to know your thoughts about this.

Best regards,
Jeanderson Barros Cândido 
http://jeandersonbc.github.io


_______________________________________________
platform-ui-dev mailing list
platform-ui-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/platform-ui-dev
_______________________________________________
platform-ui-dev mailing list
platform-ui-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe
from this list, visit
https://dev.eclipse.org/mailman/listinfo/platform-ui-dev