Description Joe Toomey 2007-04-24 11:02:24 EDT

As a side effect of the TreeViewer.setSelection() call, all of the children of each node in the tree path are created.  [This happens in AbstractTreeViewer.internalExpand(Object, boolean):1507 and then in TreeViewer.createChildren(Widget):519.]  This behavior differs from the behavior I would get if I manually expanded each node (where the tree would only create the nodes that would be visible in the viewer.)

I can envision two ways in which this could be improved:

1) In AbstractTreeViewer.setSelectionToWidget(), SWT is walking back up the specified TreePath, recursively exposing the children of each of the parent nodes.  It seems like it should be possible to optimize this traversal to stop expanding each necessary node has been found?  This would still not be perfect, since, if the node in question was at the very end of the tree traversal, we'd still end up constructing almost all of the nodes.  But in our most common cases, the user is navigating to the first element of a given type, and that element may be near the beginning of the traversal.

2) A much more efficient way would be for me to pass you the indices of the nodes in the TreePath instead of objects that you then have to search for.  I can determine the indices in question efficiently, and if I could pass you a path of indices instead of objects (or perhaps in addition to the existing TreePath), I expect you would be able to create only the necessary nodes.  This would be ideal for us.

I was able to work around this by manually forcing the creation of each node in the tree path before setting the selection (thanks for the suggestion, Boris.)