In this talk we demonstrate
Dunnart\footnote{\url{http://www.csse.monash.edu.au/~mwybrow/dunnart/}},
a constraint-based network diagram editor. In particular, we show
Dunnart's ability to browse large graphs, such as those used by the
bioinformatics community.

In Dunnart, constrained network layout is run continuously during
interaction. It constantly improves the layout while the user moves or
resizes nodes. In this model, constraints are used to maintain
non-overlap and preserve the topology of the graph during interaction,
preventing shapes from crossing shapes or connectors. The user is able to
pause the layout and make manual changes to the topology by holding down
the ALT key.  Dunnart offers the user constraint-based placement tools
for alignment, distribution or separation. These create persistent
relationships that will be enforced throughout further editing. Such
tools, are useful for drawing attention to important sections of the
diagram, or can allow parts of the layout to be constrained to match
recognised textbook biological layouts.

We demonstrate how the continuous network layout model can be
interactively used for exploration of large graphs, showing examples from
the biological sciences and software engineering domains. We use a
overview+detail model where a fast mostly-unconstrained force-directed
layout method is used to layout the entire large network in an overview
window. A detailed view shows a dynamic subnetwork of around forty nodes.
This detailed view is arranged with our standard topology-preserving
constrained graph layout.  One of the nodes is considered the active node
and we determine a surrounding network, based on graph theoretic
distance. Nodes connected (via edges) to parts of the graph not shown in
the detailed view are displayed with spikes to indicate the connections
to unshown sections of the graph. The user can middle-click on a node to
make it the focal node. This will centre the detailed view on it and
bring into view the network around it.

Once nodes have been shown in the detailed view, their positions are
subsequently locked in the overview window during relayout. Their
topology and any placement relationships they are involved in are
remembered and restored when the user brings them back into view.  In
this way the user is able to easily move throughout a large network and
incrementally tidy and beautify it while exploring.