Improved Circular Layout in IBM ILOG JViews 8.7
Below is an example graph, that is not yet laid out. Such a representation is not useful since the graph is messy, has many crossings and you cannot recognize any structure inside the graph. In previous versions of JViews, you would now need to assign cluster ids to each individual node before you can even experiment with how the circular layout might arrange the graph.
In IBM ILOG JViews 8.7, this situation has greatly improved. The circular layout got a new "automatic clustering mode". It can now determine the clusters purely from the topology of the graph, and cluster ids no longer need to be specified. The automatic clustering mode even works when the cluster ids are partially specified, that is, when you have a mix of nodes that have cluster ids and nodes that don't yet belong to any cluster. The algorithm that we use determines the clusters by detecting parts of the graph that have certain topological properties. For example, portions of the graph that are stronger connected or that represent a star shape with its incident nodes. After the clusters are automatically detected, the nodes are ordered within each cluster to avoid link crossings. All of this is fully automatic. All you need to do is to load a graph and hit the layout button, making it much easier to use than before. After hitting the layout button, the messy graph above gets analyzed and below we see the result of the circular layout: