For the purpose of fixing Java refactoring-related code to treat reduced tree mode naturally, it would be helpful to have an explicit representation of a segment of a tree:

src/
src/org/
src/org/here/
src/org/here/Main.java
src/org/here/api/
src/org/here/api/Api.java
src/org/here/internal/
src/org/here/internal/Impl.java

DataFolder adequately represents src/org/here/ and all its recursive contents; NonRecursiveFolder represents src/org/here/* (i.e. just Main.java). To treat the org.here subtree of src/, the DataFolder works for most purposes (i.e. searching, or adding things) but fails for a few purposes (moving or deleting).

Could perhaps have something like the following in a Node.lookup:

public interface FolderSegment {
  FileObject top();
  FileObject bottom();
}

with the invariants

bottom.isFolder()
top==bottom || FileUtil.isParentOf(top, bottom)
∀ FileObject f: FileUtil.isParentOf(top, f) → f==bottom || FileUtil.isParentOf(f, bottom) || FileUtil.isParentOf(bottom, f)

Thus top=src/org/ and bottom=src/org/here/ is a legitimate segment, but top=src/org/ and bottom=src/org/here/api/ is not (pick f=src/org/here/Main.java for example).

More descriptive terminology welcome; I am not sure if the above concept has a standard name.

The node would typically include bottom (and/or DataFolder.findFolder(bottom)) in its lookup as well, for callers which do not care about the top and need not use this API.