- First it separates the graph in a grid of squares.
- Then it gets which nodes are on which squares (of course a node could be on several squares, especially if it's big)
- Then, it gets a list of "proximity relations", which approximate if two nodes are in the same area. Two nodes have a "proximity relation" if they are on a common square...
- Then it tests each of these relations to eliminate nodes that actually do not overlap
- And it applies a repulsive force between the nodes that overlap.
Noverlap is then very slow if many nodes are in the same square (many proximity relations).
I'll try to put this setting (grid precision) as an editable parameter.
Nevertheless, you can improve performance by avoiding isolated nodes far away from the rest of the graph. I propose you a very ugly drawing to explain the problem...
Statistics:Posted by jacomyma — 08 Feb 2011 15:59
]]>