The layout algorithm role is to compute the best possible positions of nodes in a given space (2D or 3D) and eventually break points for edges if supported using either aesthetic constraints, hierarchical constraints or grouping constraints. As there are many such algorithms with distinct qualities and uses, this interface defines what is awaited from a general layout algorithm.
This algorithm is a {@link Pipe} that receives notifications on the grapheventually maintain an internal representation of it (or for some of them work directly on the graph), and in return send graph events to give each node a position via "xyz" attributes. Some algorithms may also export more information for nodes and edges. For example some algorithms are also able to work on the shape of an edge or the shape of a node.
The layout algorithm described by this interface may be iterative. Some algorithm will compute directly their final representation of the graph in one pass. However most algorithms will probably work step by step until a global quality function is satisfied. This is the best way to handle evolving graphs.
This behavior has been chosen because this algorithm is often run aside the main thread that works on the graph. We want a thread to be able to compute a new layout on its side, without disturbing the main algorithm run on the graph. See the {@link org.graphstream.ui.layout.LayoutRunner} for an helperclass allowing to create such a thread.
To be notified of the layout changes dynamically, you must register as a sink of the layout.
The graph viewers in the UI package often use a layout algorithm to present graphs on screen.
@since 20050706 | |