UiBinder is awesome!
An extra degree of decoupling, could be done by adding the next stuff
to the UiBinder interface:
public interface UiBinder<U, O> {
U createAndBindUi(O owner);
public static interface Pair<U, O>{
R getRoot();
O getOwner();
}
Pair<U, O> createAndBindUi();
}
where createAndBind() without parameters could instantiate an owner
class, enabling injection by constructor for UiField for more robust
code.
- Andrés
On 5 ago, 10:32, Ray Ryan <[email protected]> wrote:
> The GWT standard widgets already have custom parsers built for them.
> DockPanel, DisclosurePanel, Menubar, etc. all work just fine despite their
> wonky api needs. The bug here is that you can't indulge in similar glue for
> your custom widgets yet. No Google team has found that to be a problem with
> their custom widgets, which is I why I felt like I could get away with
> ducking that issue a bit longer.
>
> On Wed, Aug 5, 2009 at 1:31 AM, brett.wooldridge <[email protected]
>
> > wrote:
>
> > Just a question, and a comment. First the comment. Thank you for
> > getting this up into the repo, in whatever state. Second, it was
> > commented that Adwords and a few other projects have vetted this over
> > the past year. How does this jibe with the deficiencies outlined?
> > For example, not being able to markup for DockPanel, etc? Did those
> > projects just have to go "off-roading" and create custom parsers based
> > on an API they knew they would eventually have to fix/rewrite?
>
> > -Brett
>
> > On Aug 5, 5:49 am, Ray Ryan <[email protected]> wrote:
> > > I share your concern, Amir, but I'm even more afraid of a) providing an
> > ill
> > > considered API for custom parsers and b) delaying 2.0. I'm pretty
> > confident
> > > we can limp along without them for a dot release.
>
> > > On Tue, Aug 4, 2009 at 1:36 PM, Amir Kashani <[email protected]>
> > wrote:
> > > > As Ray mentioned, one has a pretty simple workaround and two is pretty
> > > > uncommon. I'm a little more concerned about the third case. A few
> > examples
> > > > of issue with internally used widgets I've created:
>
> > > > - A StackPanel replacement that adds animation support. The only
> > workaround
> > > > I can think of is having the add() method take a StackPanelItem or
> > similar
> > > > that contains the header text or widget.
> > > > - TitledPanel, which supports a header, content and footer area. In
> > this
> > > > case, the widget could expect several calls to add, and determine the
> > > > context based on number of previous calls. This would get a bit hairy
> > if
> > > > headers and footers were optional, though.
>
> > > > These scenarios are a bit inconvenient without a custom parser, but far
> > > > from a deal breaker. The concern is that people develop a set of
> > hackish
> > > > workarounds that aren't easily fixed when custom parsers are supported.
>
> > > > - Amir
>
> > > > On Tue, Aug 4, 2009 at 12:47 PM, Joel Webber <[email protected]> wrote:
>
> > > >> There are three cases where custom parsers come up:1. Widgets without
> > a
> > > >> default constructor.
> > > >> 2. Non-widget UIObjects that need an XML representation.
> > > >> 3. Panels that need more than the default add() method to deal
> > properly
> > > >> with child widgets.
>
> > > >> The former is usually pretty easy to work around, and it seldom comes
> > up
> > > >> much in practice (I think it came up for MenuBar, because it wants its
> > > >> 'direction' as an invariant -- that wasn't even a good design anyway).
>
> > > >> The second case doesn't come up all that often, but it's important for
> > > >> menus and trees.
>
> > > >> The third case is the most problematic. Take DockPanel, for example:
> > It's
> > > >> really not going to be able to do anything useful if you just call
> > add() on
> > > >> it, because it doesn't know where to put the child. These sorts of
> > panels
> > > >> need extra attributes or elements to specify where to put children.
>
> > > >> On Tue, Aug 4, 2009 at 3:42 PM, Amir Kashani <[email protected]
> > >wrote:
>
> > > >>> What are the limitations for a Widget developer without a custom
> > parser?
> > > >>> I've only begun to look at the code, but it seems like it'll still be
> > > >>> possible to use a custom widget albeit with cumbersome markup.
> > > >>> - Amir
>
> > > >>> On Tue, Aug 4, 2009 at 12:35 PM, Joel Webber <[email protected]>
> > wrote:
>
> > > >>>> Ok, then we'll need to be pretty clear about that in the
> > documentation,
> > > >>>> because it's a pretty serious landmine (i.e., in that existing
> > projects
> > > >>>> could easily have some widgets that couldn't be directly used with
> > UiBinder
> > > >>>> without hackery). As an example, I'm going to have to add some
> > parsers for
> > > >>>> LayoutPanel, et al, because they have somewhat unusual construction
> > > >>>> semantics.
>
> > > >>>> On Tue, Aug 4, 2009 at 3:31 PM, Ray Ryan <[email protected]> wrote:
>
> > > >>>>> I was thinking 2.1, actually.
>
> > > >>>>> On Tue, Aug 4, 2009 at 12:31 PM, <[email protected]> wrote:
>
> > > >>>>>> On 2009/08/04 18:50:55, Ray Ryan wrote:
>
> > > >>>>>>> On 2009/08/04 17:44:38, Ray Ryan wrote:
>
> > > >>>>>> A question for the group: the stuff under rebind and parsers
> > should
>
> > > >>>>>> not be
>
> > > >>>>>>> considered public API, it's just not ready for that. Is javadoc
> > to
>
> > > >>>>>> that effect
>
> > > >>>>>>> enough of a deterrent? (Although I suppose the fact that you
> > can't
>
> > > >>>>>> actually make
>
> > > >>>>>>> your own parsers and such *do* anything yet will make the issue
> > > >>>>>>> moot.)
>
> > > >>>>>> I would assume that if you can't usefully write your own yet, then
> > > >>>>>> it's
> > > >>>>>> pretty safe to keep evolving the API. I assume that there's a
> > > >>>>>> 2.0-time-frame task to make a public API for parsers?
>
> > > >>>>>>http://gwt-code-reviews.appspot.com/51831
--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---
