Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [eclipse-incubator-e4-dev] Declarative UI

Hi folks,

I just got caught up on this great thread.

Hallvard, good summary.

As another cut through the space, this is what I'm hearing, two extremes, both compelling:

A) A "literal" model of the UI.
You know the target is SWT and the model reflects exactly that.  If you know SWT, you can guess pretty easily what to put in the angle brackets.  You get a serialization/deserialization format.  For large UIs it can reduce your total Java code in-memory footprint.  You can do things like build up the model in memory first, then create the widget structure as a walk through the model.  You can build that model from reading the current widget tree.  It allows you to avoid widget reparenting issues by reading the widgets, destroying them, then recreating them elsewhere (e.g in a second workbench window).  It facilitates GUI builders.  Its a good basis from which to build other model abstractions (see below).

B) An abstract model of UIs (say, functional).

Here you want to get "above" the level of the widgets.  The domain of discorse is maybe one of data display and user interaction, or common UI solution patterns, not widgets and layouts.  This might be a model of UI requirements and capabilities that get translated into concrete widgets. For example, "I need a table, the user will add and remove from the table" generates a table widget with add/remove buttons.  This can be handy if you are generating a lot of UI panels for say pages of entry fields, dialogs, and wizards.  It also provides a natural way to bridge to different widget sets (e.g. gen to SWT, gen to web).  You likely forgo some ability to tweak the UI but gain in speed of authoring, easier refactoring, etc.  You also get free consistency in your UI since functionally equivalent UI areas get gen'd the same.  Such models are much more contentious though since you must convince yourself you're modelling the right thing, and that either the limitations of the model are acceptable or there's some "escape to assembler" path.

These two aren't mutually exclusive.  If you're doing (B), its probably handy having (A) so you can do a model to model transformation as the path to instantiating the actual widgets.

The practical me likes (A) since the benefits are clear and you don't need to think too deeply about the model (this is a feature).
The theoretical me likes (B) since I really love the notion of elevating the task of programming beyond the mostly administrative task of 3GL programming.

I need to learn more, but it sounds like XSWT is close to (A) although you might not get everything currently like the live model.
Again I need to learn more, but is XAML much more than (A)?  It seems less than (B) since I'm still talking about a literal UI with a certain layout, labels, etc., although perhaps in a more general manner.


"Yves YANG" <yves.yang@xxxxxxxxxxx>
Sent by: eclipse-incubator-e4-dev-bounces@xxxxxxxxxxx

11/09/2008 11:16 AM

Please respond to
E4 developer list <eclipse-incubator-e4-dev@xxxxxxxxxxx>

"'E4 developer list'" <eclipse-incubator-e4-dev@xxxxxxxxxxx>
RE: [eclipse-incubator-e4-dev] Declarative UI

Hi Hallvard,

Regarding the comparison between XAML and XSWT, I agree with you.

As for the goals, I think the goal (1) is a wrong goal for XML since there
are other formats are more compact like JSON, or binary directly. The goal
(2) is necessary as declarative UI language. The goal (3) is rather than a
model in memory. I vote (2) + (3).

-----Original Message-----
From: eclipse-incubator-e4-dev-bounces@xxxxxxxxxxx
[mailto:eclipse-incubator-e4-dev-bounces@xxxxxxxxxxx] On Behalf Of Hallvard
Sent: Sunday, November 09, 2008 4:54 PM
To: E4 developer list
Subject: Re: [eclipse-incubator-e4-dev] Declarative UI


Kevin McGuire wrote:

> Shall I try to pick a time in say about two weeks?

OK for me. It has been mentioned that some will meet at ESE. I'm going there

myself, but haven't yet received a confirmation about the e4 symposium. I
was a
day late with the position paper (didn't post here but to the email address
found in the call), but still I hope to be allowed to attend.

> Specifically, while it would be good to have presentations on XSWT, XUL,
> I think the agenda has to be more around approaches and goals.

In the last postings (Yves Yang and David Orme), there has been some
of XML notations (XAML and XSWT). As pointed out by Yang, XSWT is perhaps
targeted at serializing SWT. Or rather, it is a description of how to
(procedurally) _build_ a specific SWT UI, while a XAML is a (declarative)
of the (initial) _state_ of a UI. David and Yves, do you agree? Whether or
this difference is important, depends on the goal.

If the goal (1) is a compact notation from which to render an SWT UI, I
XSWT's approach (guessing the meaning of names, based on type information
acquired by means of reflection) is best. If the goal (2) is a model for
describing the state of the UI, a model which lends itself to generation,
transformation and other processing, something XAML-like may be best. A
(level) goal (3) is having a "live" model which is kept in sync with the
UI. Such a model can provide a more uniform, API-independent way of
 and listening to the UI.

These are different goals I can identify, and that I hope is a constructive
contribution to this discussion.

Personally, I favor goal (3), a model supporting live syncing with the real
If we design this model using EMF, we get scripting and databinding for free
already exists). I think we can still use XSWT, XAML and other XML languages
concrete syntaxes, since they may be more usable than EMF's native one. The
processor can we thought of as a compiler that generates the EMF model as
intermediate code, instead of building the SWT UI directly. It should also
possible to generate Java code from the EMF model, or perhaps compile it
directly to bytecode, for use in mobile applications.

eclipse-incubator-e4-dev mailing list

eclipse-incubator-e4-dev mailing list

Back to the top