[jbossseam-issues] [JBoss JIRA] Updated: (JBSEAM-3645) Support co-existence of multiple view layers in same deployment

Support co-existence of multiple view layers in same deployment --------------------------------------------------------------- Key: JBSEAM-3645 URL: https://jira.jboss.org/jira/browse/JBSEAM-3645 Project: Seam Issue Type: Patch Reporter: Clint Popetz Assignee: Clint Popetz Attachments: multi-view-take-2.patch, multi-view-take-3.patch, multi-view.diff Currently one must choose between wicket and jsf when using seam, because the wicket support works by overriding certain faces components. This patch eliminates that limitation. This patch should be applied after the patches from the issues linked to this issue, which are also related to wicket development. There are four parts of this patch. (1) The Manager and StatusMessages components are currently implemented in jsf with FacesManager and FacesMessages, and in wicket with WicketManager and WicketStatusMessages. Since a mixed deployment would need all managers to be @Installed, this patch makes each have a separate @Name, and makes the base classes (Manager and StatusMessages) @Unwrap themselves as one of the subclass components. Each subclass component registers itself with the base class component (by observing the postInitialization event) as a possible implementation. When Manager.instance() or StatusMessages.instance() is invoked, and the base component is created, @Unwrap is called, and that method asks each subclass component whether they should be used for that event scope. The Wicket implementations say yes if the wicket Application.exists() (which uses a ThreadLocal) returns true. The Faces implementations says yes if facesContext is not null. This patch assumes that both will not simultaneously return true, and if so, the behavior is undefined, as it's first-come-first-serve. If neither return true, the default Manager and StatusMessages implementation will be used. This happens during initialization and other edge cases where a manager is needed for event contexts that aren't tied to real requests. Note that since the FacesMessages component no longer has the same @Name as the StatusMessages component, it is now "org.jboss.seam.faces.facesMessages" which means the two factories for that name and its unqualified form in jboss-seam.jar's components.xml have been removed. (2) Seam's WicketFilter currently sets up scopes unconditionally if it is installed. This causes faces requests to fail because the SeamPhaseListener also sets up and tears down scopes, and so when the wicket filter finishes and tries to tear down its scopes, it finds them gone and an exception occurs. I've changed WicketFilter to not set up scopes, but instead to have WicketFilterInstantiator's anonymous subclass of the real WicketFilter install the scopes _if_ wicket has determined the request is a wicket request. (3) WicketExceptionFilter existed only to turn off the base ExceptionFilter for wicket classes. I've removed this class, and we'll need to add to the wicket seam documentation the need to map the ExceptionFilter in components.xml to jsf paths if wicket is to be used alongside jsf. (4) WicketRedirectFilter now co-exists with the base RedirectFilter, rather than replacing it. It checks to see if Wicket is active (via the same Application.exists() call as above) before doing any work to redirects, so that it doesn't duplicate work that the jsf RedirectFilter does. Likewise, The jsf RedirectFilter now checks to see if there is a faces context before doing any work.