af:tree leaf nodes

Question

Is there a way to display leaf nodes in an <af:tree> differently, so that they don't look like they can be expanded?
I'm using an <af:tree> component with an ADF tree binding. There's just one source collection, with a self-join (that is, the single accessor rule returns different elements from that collection).
The problem is that every node in the tree is expandable, including the leaf nodes (that is, the ones with no children). When the user expands them, nothing happens except a change in the icon (as you might expect), but I'd rather not provide the icon at all, or provide a different icon, to discourage expansion.
I actually have an attribute in my collection that can act as a discriminator--I happen to know that all and only rows with TypeIndicator="W" will have children. But my attempts to use the "polymorphic rule" functionality in the tree binding editor don't seem to work--it simply doesn't show any children of the top-level nodes at all.
Help would be much appreciated.

Answer 1

Mmm...sorta. Here's what I've done so far.
I've created a class, let's call it CustomTreeModel, that extends the (abstract) class TreeModel. I've given it a private field:
private final TreeModel mTree;and a constructor:
public CustomTreeModel(TreeModel tree) {
    mTree = tree;
}Then, I've overridden every method but one in TreeModel to delegate to mTree. For example:
public void setRowIndex(int i) {
    mTree.setRowIndex(i);
}The one exception is the method isContainer, which I've implemented this way:
public boolean isContainer() {
    return (mTree.isContainer() && (! mTree.isContainerEmpty()));
}The idea here is that the class wraps a tree, except that it claims that empty containers aren't containers at all (so they should be treated as leaves).
Then, rather than bind my <af:tree> to #{bindings.MyTreeBinding.treeModel}, I bound it to #{myBackingBean.treeModelWrapper}, and created a backing bean with bean name myBackingBean. Here's what I did in the backing bean:
1) Added a managed property, bindings, of type BindingContainer and value #{bindings}, and added a BindingContainer field, bindings, to the backing bean class.
2) In the bean class, added a TreeModel field, treeModelWrapper.
3) Changed the getter for treeModelWrapper so it looked like this (this is inexact, I don't have access to my code right now).
public TreeModel getTreeModelWrapper() {
    Map myTreeBinding = (Map) getBindings().getControlBinding("MyTreeBinding");
    TreeModel oldTreeModel = (TreeModel) myTreeBinding.get("treeModel");
    return new CustomTreeModel(oldTreeModel);
}That worked, inasmuch as the general structure of the tree looked about right, and leaf nodes showed up as leaf nodes.
There are two problems with it. A little one and a big one.
Little: It degrades performance significantly. Not surprising, since I assume isContainerEmpty() is a more resource-intensive method than isContainer(), and we're calling it a lot more often now. There's probably no way around this.
Big: The data seems...a bit messed up. Most of it looks right, but I'm getting weird occasional spurious child nodes (meaning that occasionally a copy of a node is showing up in a container where it doesn't belong). Since the tree is based on a self-referential relationship, that occasionally leads to infinite loops (node A and node B show up inside one another, so you can just keep expanding downwards for ever). I can't for the life of me figure out why this is.