Actually, I think the code for this late-bound lazy model that saves space
is fairly straightforward. Something like (notice no assignment to
this.model,
so the wrapper is recreated every time):
public IModel getModel()
{
// If model is null
if (model == null)
{
// always recreate model wrapper for late-bound models
to save space
return initModel();
}
return model;
}
Another maybe better option would be to null out any models that were
created
by initModel on detach(). Yet another boolean to do that, but it's more
space
efficient and doesn't always recreate the model wrapper either... assuming
we've
got the extra bit still, which at first glance it looks like we do... this
seems like
the way to go...
Jonathan Locke wrote:
>
>
>
> Johan Compagner wrote:
>>
>> It could be protected if you like. (maybe final then?)
>> But it is a very simple method that you can implement very easy in your
>> validator self ofcourse.
>> (then you also don't have a back call to component)
>>
>
> I don't see how you can do that. Validators no longer know about models.
> That's all hidden behind IValidatable.getValue() now. I think that's
> good,
> because they are more general now, but it's a limitation for this use
> case.
>
> The way I have implemented this now, the validator is not aware of any of
> this.
> It has a constrainedClass property, which is set for
> CompoundPropertyModels
> when the component attaches. This is a bit ugly, but necessary because
> there's
> no way through the validation API to get the model, only the object to be
> validated,
> and there's no way in the constrained field to get the class of the model
> because
> that may be unknown until the compound property model binding occurs via
> initModel(). Make sense? Simplest thing is to make getRootModel(IModel)
> protected
> and maybe implement a public getRootModel() for convenience that does
> getRootModel(getModel()), which is what I'm doing.
>
>
> Johan Compagner wrote:
>>
>> No not really, generics and the previous compound don't really work
>> together
>> correctly and then we need the Component param back in the normal model
>> which is ugly and for the compound param very ambiguous what to put in
>> it.
>> currently component.getModelObject() is always
>> component.getModel().getObject()
>>
>
> I didn't mean we should have a component parameter. I was wondering more
> if the wrapper couldn't be transient (construct a wrapper every time we're
> asked
> for the model and then let the wrapper get GC'ed rather than holding on to
> it.
> if that's possible, it could mean memory savings, but also getting rid of
> that
> interface too)
>
>
--
View this message in context:
http://www.nabble.com/getRootModel-and-tf3476307.html#a9708213
Sent from the Wicket - Dev mailing list archive at Nabble.com.
