[Jung-support] jung comments

[Dave A. <dc...@mi...>]

We're continuing our evaluations, and generally quite happy with Jung 
overall.
I have a couple of specific pieces of feedback that you are welcome to take
for what they're worth.

First, we saw a performance scalability issue that may be quite easy to 
resolve.
This is on the order of 10,000 vertices and 100,000 edges.  Constructing a
graph (separate from visualizing it) took quite a long time, and a 
substantial
chunk of that was due to the duplicate edge checking.  Specifically, the 
fact
that the AbstractSparseGraph maintains HashMaps of edges and nodes
keyed by Integer objects denoting the edge and node IDs, while the edges
and nodes return ints for their ids.  This means every addition (and 
many other
operations) causes a "new Integer()" construction.  Would it make more sense
to use the ID as the object's hashCode and use a HashSet, and eliminate the
Integer construction entirely?

Second, an architectural comment.  The user interaction function appears
to be VisualizationViewer.GraphMouse, which can be replaced by overriding
initMouseClicker() in a subclass.  This isn't too onerous, but you might 
want
to more explicitly expose how the MVC Controller component (GraphMouse)
can be replaced.

-dca