-Igor
On 11/28/05, Eelco Hillenius <[EMAIL PROTECTED]> wrote:
I just fixed the ajax implementation. Improved it greatly actually.
Because of the new IRequestTarget architecture, AjaxHandlers are now
totally free to choose their own way of responding elegantly (instead
of the hackish backdoor approach of rendering directly to the response
stream). The default way is that they still work with a
IResourceStream (as provided by getrResponse), but ajax handler
implementations can set whatever request target they want. For
instance you should now be able to render a single component ( e.g. a
panel) back by setting RequestCycle.setRequestTarget(new
wicket.request.ComponentRequestTarget( ... component ... ));
Have fun exporing!
Eelco
On 11/28/05, Eelco Hillenius < [EMAIL PROTECTED]> wrote:
> Oh, I didn't have time yet to test the ajax behaviour. I could use
> some help there. I'm about to close down here (or sleep). Anyone in
> Europe (Marco, Ruud, Johan, Martijn, ...) want to test and possibly
> fix this? Thanks,
>
> Eelco
>
>
> On 11/28/05, Eelco Hillenius <[EMAIL PROTECTED]> wrote:
> > Hi all,
> >
> > I've been working on some heavy duty refactorings of RequestCycle the
> > last few days. I'm far from done, but I just reached an important
> > milestone in it and I thought it was time to explain a bit what I am
> > doing for those that follow HEAD.
> >
> > There's a TODO document in the new niceurl example of wicket-examples.
> > Juergen, if you're reading this, could you take a look at that?
> > There's a question for you in it :).
> >
> > Here's the last commit log in which I try to explain the refactorings:
> >
> > ...
> > work in progress: request cycle refactoring.
> >
> > Milestone: the major part of low-level refactoring is done now and
> > there is new functionality
> > for mounting request targets to paths.
> >
> > A short explanation how request processing works now:
> >
> > Request cycle processing is now mostly delegated to an instance of
> > IRequestCycleProcessor which can
> > be created by RequestCycles or - preferably - the default request
> > cycle processor in Application.
> >
> > I defined the processing clear steps; they can be found as methods on
> > IRequestCycleProcessor:
> >
> > 1. Destiling a strongly typed RequestParameters object from the
> > Request. This is done by an instance
> > of IRequestEncoder. The request encoder is the strategy that takes
> > care of creating urls (all urlFor
> > methods of Page are now dispatched to this strategy) and 'decoding'
> > them into the strongly typed
> > parameters object. Furthermore, the encoder interface provides for
> > mounting and unmounting
> > Wicket request targets (see 2). Using this interface has the
> > advantage that all 'translations' are now
> > being done by one object and that the translations do not do any real
> > processing yet. Hence the
> > actual processing can be implemented in a very different way that has
> > nothing to do with url translations.
> >
> >
> >
> > 2. Solving an incomming request to a request target using the strongly
> > typed RequestParameters of step 1.
> > Request targets (IRequestTarget) are a central concept in the new
> > refactorings. A request cycle has
> > one current request target that can range from a bookmarkable page
> > (or a redirect to it), a component
> > listener invocation to shared resources, external resources or
> > something like a HTTP error message.
> > Request targets themselves are responsible for creating/ delegating a
> > response. For the rest they are
> > fairly shallow, and implementations of IRequestCycleProcessor do the
> > heavy lifting.
> >
> > Once the target is determined, it is set as 'the current' target in
> > the RequestCycle. Actually, RequestCycle
> > holds a stack of all set targets. For instance, when you call
> > setResponsePage, what happens is that
> > actually a request target implementation (in this case
> > (PageRequestTarget) is put on top of the stack.
> > after the request cycle is done executing, all request targets will
> > be iterated and their cleanUp methods
> > will be called to enable request target implementations to free
> > resources, give notifications, etc.
> >
> > 3. The next step is event dispatching. During this step things like
> > calling listener interface methods (such
> > as Link.onClick) may be done.
> >
> > 4. The last step is to respond using the current request target. A
> > target may directly call render on e.g.
> > a component, delegate the actual response creation (like with
> > resources) or create a redirect to
> > defer the actual handling. A target implementation is free to do
> > whatever it wants, as long as a response
> > is created.
> >
> > 4.a. when something during handling of 3 or 4 goes wrong, an exception
> > response is created. It is now much
> > easier (I hope) for users to implement completely custom exception
> > handling/ page rendering.
> >
> > NOTE: most of the refactorings are interface based (due to popular
> > demand?) so that it is much easier than
> > before to create proxies, interleave them, etc. This can now be done
> > better than before as the contract
> > is better externalized now (and enforced in the final method
> > RequestCycle.request()) and devided in seperate
> > steps instead of being implementation details.
> >
> > NOTE: there is a lot of cleaning up to do still. I'd like to take a
> > good look at all the special cases we put in
> > Application/Settings/Pages
> > the last few months, and get rid of the things we don't need to support anymore
> > when they can be done using this interface, such as exception page handling.
> >
> >
> > NOTE: there is a default, handy implementation of
> > IRequestCycleProcessor called CompoundRequestCycleProcessor
> > which delegates it's behaviour into seperate interface calls so that
> > it is easier to mix-'n-match implementations.
> >
> > NOTE: there are some API breaks:
> >
> > - Page class alias functionality is not supported anymore. It was
> > 'miss-used' for shortening resource keys
> > more than it was for having page aliases, and while the shortening of
> > resource keys might be a nice thing
> > to have again in the near future, the page aliases are not needed
> > anymore as it is replaced by the new
> > path mounting functionality. We need to rethink the shortening of
> > resource keys in a more generalized
> > fashion in the next few days.
> >
> > - The home page render strategy is removed. There are now two options:
> > 1. The home page class is mounted to a path. When that is the case, a
> > redirect to this bookmarkable page
> > path will be issued and the mounted url is thus shown.
> > 2. Nothing was mounted. The home page is rendered directly and the
> > url will be 'clean'
> > (like http://localhost:8080/wicket-examples/navomatic)
> >
> > I know there was a third option, but it was not clear to me what the
> > advantage for that one was. If someone
> > needs it, please complain :)
> >
> > ...
> >
> > Hope you have fun with it. Tomorrow, I'll (finally, sorry Evan)
> > implement browser detection, which should be fairly simple now.
> >
> > Eelco
> >
>
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&opclick
_______________________________________________
Wicket-develop mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wicket-develop
