I believe this is not possible under vanilla JSF 1.1. It's my
understanding that the component tree is build as the components are
rendered.
Under JSF 1.2 as well as under JSF 1.1 + facelets, the component tree
is build first, then the components are rendered. So it was a known
design flaw of JSF 1.1.
On 4/12/06, Rhys Parry <[EMAIL PROTECTED]> wrote:
> All,
>
> I am having a problem in achieving the behaviour I desire for rendering
> responses. I would like to, at PhaseId.RENDER_RESPONSE, iterate the
> component tree, and based on clientId setRendered(true | false). The problem
> is that I never get the component tree. It seems that the tree is created
> AFTER the PhaseId.RENDER_RESPONSE.
>
> I have the folowing code:
> public void beforePhase(PhaseEvent pe)
> {
> UIViewRoot root=pe.getFacesContext().getViewRoot();
> log.debug( "root name = " + root.getViewId());
> log.debug( "child count = " + root.getChildCount());
> log.debug( "child count2= " +root.getChildren().size());
> log.debug( "facet count = " +root.getFacets().size());
> iterateComponentTree(root.getFacetsAndChildren(),
> pe.getFacesContext(), 0);
> }
> where iterateComponentTree() is a recursive method unpacking the tree and
> looking at the permissions for a component vs. logged in user and setting the
> components setRendered(true | false). The problem is that there is no
> component tree. The call root.getChildCount() returns 0. I do not want to
> write rendered="#{someBean.wasteOfCode}" on every component.
>
> I have tried the same code in afterPhase(PhaseEvent) and am able to retreive
> the tree but cannot effect its rendering behaviour.
>
> Perhaps someone knows of a hook that I can use?????
>
> If not then:
> What I would like to see is a configurable option to inject a RenderManager
> type bean that subscribes to some interface with one method taking the
> clientId as the param. I could then take the clientId passed to the
> RenderManager and map my security concerns to it. UIComponentBase could use a
> simple class like this in its isRendered method.
>
> It seems that my only option at this point ( that I am aware of ) is to use
> AOP.
>
> Hope to hear from someone,
> Rhys
>
