Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [e4-dev] Declarative UI in E4


In wazaabi 2.0, I choose to use 2 metamodels: one is abstract and holds the behavior which is not SWT specific. The second one is concrete and tries (with respect to the abstract one) to be as close as possible to SWT. This has been done for allowing other concrete models like Swing (which is not one of the first target of E4, I know). The runtime engine is split up into 2 parts respecting the separation of metamodels. The mechanism is based on controllers which manage views (1 model <->1 controller <->1 view) and the view contains usually a widget. The view creation is not dependant of the parent, even if the underlying SWT widget is.
It allows the reparent of any view and/or provides skinning capabilities.
This separation allows also stylesheet mechanism (not implemented yet). (A button could be systematically embedded in a container)

IMO, having a 1:1 mapping means we just reproduce the implicit model of the APIn then we accept as a model langages constraints. For instance, for the checkbox button, is the selection (check) an event or a property ? In wazaabi it acts as a property (and of course, EMF allows us to add Adapter on it). My current work is focused on databinding model & engine. After this (important step), I will try to build E4 demo using wazaabi.



Kevin McGuire a écrit :

My impression was that XWT and Hallvard's model were modelling at two different levels. At least, that's the interesting part to me (and less so, it's representation as an EMF model or XML). I admit I've not looked closely at either due to lack of time (sigh), but at a broad level, I believe that Yves approach maps more closely to the widgets but is not a direct representation of them (close but not 1:1), while as Hallvard's is a more abstract approach which could facilitate higher level UI expressions, eventually generated down to widgets (even less 1:1).

I'm not sure I've characterized them accurately and if not apologies.

I think both approaches are interested and valid. A close mapping to the widget structure allows for GUI builders, serialization/deserialization, testers, etc. A more abstract model might allow you raise the level of expression of the UI to more functional needs ("I need a list of directories" vs. "I need a list widget whose content provider gives me directories").

As a super-minor example, in CSS you can set margins. Our SWT widgets don't have equivalent API for margins. The way to implement it in SWT is with a Composite and layout. But my declarative UI likely won't have that composite. Ideally the widget generation would be in cohoots with CSS and gen the right composites with layout values. So not 1:1.

*David Orme <djo@xxxxxxxxxxxxxxxxxxxxxxxxx>*
Sent by: e4-dev-bounces@xxxxxxxxxxx

01/27/2009 03:03 PM
Please respond to
E4 Project developer mailing list <e4-dev@xxxxxxxxxxx>

	E4 Project developer mailing list <e4-dev@xxxxxxxxxxx>, andy@xxxxxxxxxx
	[e4-dev] Declarative UI in E4


Two things, by way of moving the discussion forward while the IP machine grinds along:

(The impatient or overworked can read just the boldfaced text and get the gist of what I'm saying. ;-)

Thing #1:)*

My understanding is that our status is that Hallvard's code is making its way through the IP Review process and onto the E4 server.

In the meantime, I had an idea that might unify the two current proposals.

(If I mis-characterize any of this, please correct me, but here is my understanding of our current state:)

1) Hallvard's proposal is an XML layer on top of an EMF model that is connected to the UI.

2) Yyves's proposal is to map XAML 1:1 to the SWT API. He calls the resulting API XWT.

Here's what I was wondering: *Yyves, would it be possible/desirable to make XWT map 1:1 to Hallvard's EMF model rather than straight to SWT? This seems like it might be a sensible approach to unifying both proposals.*

If the answer is "yes", I'd very much like to play with that and see what the resulting code looks/feels like.

Thing #2:)*

I've been thinking about the ways we have discussed for abstracting and simplifying UI creation. So far, we have only focused on XML representation, but I recalled that Andy Maleh created (and proposed) Glimmer, a JRuby-based UI DSL project. And I've been interested in translating Glimmer's ideas into Scala.

(I noticed that Boris invited Andy to participate over here, and regrettably, he has not materialized yet. Since I know Andy personally, I'll reach out to him and invite him to join the conversation.)

In any case, *I want to make sure that whatever we do does not preclude Glimmer-like things in alternate JVM languages, whether JRuby, Scala, JavaScript, etc.*
In the abstract, it seems like it ought to be possible to map Glimmer(.JRuby | .Scala) to EMF just like Hallvard maps XML to EMF.*

Which brings me to my

Hallvard: Any ETA on when we will get to see your code at*

I'd really like to get my hands on both of these proposals and play with them. :-)

Best regards,

Dave Orme

PS: Andy: I've CC'd you on this mailing list message. If you're interested, please add yourself to the list so you can take part in the ongoing conversation. You know you're welcome here. ;-)

e4-dev mailing list


e4-dev mailing list

Internal Virus Database is out of date.
Checked by AVG - Version: 8.0.176 / Virus Database: 270.10.13/1912 - Release Date: 23/01/2009 18:54

Back to the top