I just leveraged the WicketFilter and the Factory pattern in wicket. I did
not want to just rewrite all that when the WicketFilter was already
designed to be extensible as the CdiWicketFilter does. I guess just
rewriting the createApplication in the CDIWebApplicationFactory to not
fallback to the super classes createApplication if no Applications are
found using CDI would created the all or none solution, the the other cases
in the code are useful IMO. First if you did not stress out over the extra
line of code in the WebApplication and annotated with
@WicketApp("SomeName") then the factory simply gets the instance via the
name. If the extra line of code was too much then it verifies there is one
and only one Application in the Instance<WebApplication> variable. If
there is only on WebApplication the it uses it. However if there are more
then we have an ambiguity, and fails loudly. If anything the search via
applicationClassName should be added to allow for the fully qualified name
as is the normal with the default filter.
On Mon, Nov 11, 2013 at 8:20 AM, Emond Papegaaij <[email protected]
> wrote:
> I reread the code, and I see what it is used for. CdiWebApplicationFactory
> can use a bit of a cleanup. For example, Instance has methods to check
> for satisfaction and ambiguity. Also, I think this should be rewritten as
> an
> all-or-nothing solution. You either use CdiWicketFilter with
> CdiWebApplicationFactory or you don't. This should greatly reduce the
> amount of code and should allow for more robust error checking.
>
> Best regards,
> Emond
>
>
>
> On Monday 11 November 2013 08:00:10 John Sarman wrote:
> > It is not a forced requirement, just an option for full cdi injection.
> >
> >
> > On Mon, Nov 11, 2013 at 3:50 AM, Emond Papegaaij
> <[email protected]
> > > wrote:
> > >
> > > Hi John,
> > >
> > > I've just merged the pull request in the wicket-6.x branch (still under
> > > experimental). The version still is 0.2-SNAPSHOT, as the versions are
> > > automatically increased on release. The reason I've merged the pull
> > > request is to give us all a common baseline to discuss. Could you
> please
> > > verify I did not break anything merging it? All testcases but one pass.
> > > The
> > > failing testcase (CdiConfigurationTest.testMultiAppLoad) tries to
> access
> > > the
> > > BeanManager from an unmanaged thread, resulting in an NPE.
> > >
> > > I've already noticed one aspect I do not like: the requirement to
> annotate
> > > your app with @WicketApp. With a Producer method, it should be
> possible
> > > to use the actual application names, without the requirement to
> duplicate
> > > them on your application class.
> > >
> > > Best regards,
> > > Emond
> > >
> > > On Sunday 10 November 2013 16:44:28 John Sarman wrote:
> > > > Edmond,
> > > > On July, I worked vigorously to get to the 0.3 snapshot, which was
> what
> > > > I
> > > > consider the first beta ready version of the move to cdi1.1. The 0.1
> > > > and
> > > > 0.2 snapshot was 0.1, getting it to work and learning how to
> request
> > > > pull
> > > > requests. 0.2 was adding some slight fixes and testing. After that
> I
> > > > realized that I was treating the @ApplicationScoped as same scope
> that
> > > > ThreadContext gives to a Wicket App. That is entirely wrong. So the
> > > > previous version only properly supports at most 1 Wicket app, the
> > >
> > > second
> > >
> > > > could override the Configuration of the first (not acceptable). In
> my
> > >
> > > 0.3
> > >
> > > > version, I added the code to prevent that, by using the Wicket app
> key
> > > > generated as the key to the configuration properties for an app.
> This
> > > > allows for multiple Wicket apps to be deployed in a Servlet.
> However,
> > >
> > > for
> > >
> > > > whatever reason, that checkin could not properly merge into the 7
> > >
> > > branch.
> > >
> > > > I have to remedy this even if I just have to copy paste the code, to
> > >
> > > make
> > >
> > > > git happy ( I blame myself, not Git). In the meantime, I recommend
> > >
> > > looking
> > >
> > > > at my latest (albeit broken) pull request
> > > > https://github.com/apache/wicket/pull/50 and port that version. It
> adds
> > > > thorough testing, fixes the multiple deploy issue, reintroduces the
> auto
> > > > Conversation, and extends the ConversationalComponent by
> introducing
> > >
> > > the
> > >
> > > > @Conversational, which by default works the same as the Cdi-1.0
> > > > ConverationalComponent, but also allows the propagation and auto
> > >
> > > feature to
> > >
> > > > be modified for an Object that uses the annotation, without
> affecting
> > > > the
> > > > global defaults set during Configuration. The 0.3 also introduces
> the
> > > > CdiWicketFilter. The CdiWicketFilter allows the configuration
> settings
> > >
> > > to
> > >
> > > > be managed in web.xml. It also instantiates the WicketApplication
> using
> > > > Cdi so that the Application is injected before the init() method.
> The
> > > > changes do not break the original Cdi-1.0, initialization technique,
> to
> > > > support the backwards compatibility.
> > > >
> > > > John
> > > >
> > > >
> > > >
> > > > On Sat, Nov 9, 2013 at 3:29 PM, Emond Papegaaij
> > > >
> > > > <[email protected]>wrote:
> > > > > In wicket 6, this code also still is in experimental. The reason I
> > >
> > > ported
> > >
> > > > > it to Wicket 6, was to actually use it. wicket-cdi (the old
> module),
> > > > > is
> > > > > usable with 1.1, but not very optimal. One of the main problems
> with
> > >
>
