Re: [e4-dev] E4 Extension Points

[John Arthorne <John_Arthorne@xxxxxxxxxx>]

I like option 1 as well. Generally, we
should absolutely minimize the friction needed to migrate to the "new
way". If possible, just lifting the restriction that the view must
implement IViewPart seems like the easiest migration path for plugin writers.
The flip side is that we would *only* attempt injection in the case where
the object does not implement IViewPart - I don't think we want to support
blended combinations of old API plus injection.

This has the advantage to me that we
change the schema and the tooling no longer encourages developers to do
things the "old way" at all. Of course PDE templates would also
need updating to generate new style of parts. If there are good reasons
we can't make old and new extensions work with exactly the same schema,
then some minimal change to the schema to support both cases might be possible.
We really need to experiment with this to see what works, but my general
bias is:

- All existing valid extensions must
continue to work
- Minimize migration effort for plugin
developers as much as possible
- Have schema and PDE templates encourage
the "new way" when creating new extensions

John





From:      
 Patrick Paulin <patrick@xxxxxxxxxxxx>
To:      
 E4 Project developer
mailing list <e4-dev@xxxxxxxxxxx>, 
Date:      
 07/11/2013 01:19 PM
Subject:    
   Re: [e4-dev]
E4 Extension Points
Sent by:    
   e4-dev-bounces@xxxxxxxxxxx

---

I agree with Lars that option 1 is better. But is there
a reason we couldn't specify a POJO in the normal view "class"
attribute? Whether the class implements IViewPart or is a POJO that needs
to be wrapped seems like an implementation detail. 

--- Patrick

On Jul 11, 2013, at 4:34 AM, Lars Vogel <lars.vogel@xxxxxxxxx>
wrote:

My understanding of 1.) is that the framework would accept
Pojos. In would allow a smooth migration of the existing Eclipse IDE plug-in
projects. 



2013/7/11 Jonas Helming <jonas.helming@xxxxxxxxxxxxxx>
Hi,
I agree with Lars. However, even option 1 seems like a duplication of fragments
and processors to me. The main disadvantage for me would be, that you still
stick to the old API. What is the advantage?
Regards
Jonas

Am 11.07.2013 11:22, schrieb Lars Vogel:
Hi Eric, 

I think 1.) would be the right thing. 2.) feels like a
duplication of model fragments and model processors to me.

Best regards, Lrs


2013/7/10 Eric Moffatt <emoffatt@xxxxxxxxxx>
I'm currently looking at what we're
going to do as far as extension points go to enable folks to contribute
e4 (DI) code into the IDE and I want to get some feedback from the e4 community
as to the best way for me to do this...

I have two possible approaches:

1) Extend the current IDE extension points with e4-specific sections (i.e.
extend the existing org.eclipse.ui.views EP to allow the addition of 'e4
View')
2) Provide separate extension points for the e4 bits (i.e. clone the existing
org.eclipse.ui.views EP and tweak it to be e4-related

Do you *want* e4 specific extension points ? This is independent of having
the ability to contribute them through fragment / model processing (which
we'll also be working on in Luna). The BOF at last year's eclipsecon didn't
come to a resolution on this (at least not one that I remember..;-).

Let me know what you think,
Eric


_______________________________________________
e4-dev mailing list
e4-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/e4-dev




_______________________________________________
e4-dev mailing list
e4-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/e4-dev



_______________________________________________
e4-dev mailing list
e4-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/e4-dev


_______________________________________________
e4-dev mailing list
e4-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/e4-dev



_______________________________________________
e4-dev mailing list
e4-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/e4-dev