Interface DecrossTwoLayer<Ops>

sets the initialization passes before decrossings

For every initialization operator, this will run the two layer heuristic, ultimately choosing the ordering that minimized overall crossings. For this reason, only quick decrossing operators should be used, not expensive ones. The empty list is treated as a singleton list with a noop operator.

(default: [decrossDfs(), decrossDfs().topDown(false)])

get the current initialization passes

sets the Twolayer accessor for minimizing a layer at a time

The Twolayer operator takes pairs of layers, and reorders one. There are three built-in variants:

• twolayerGreedy - This takes another two-layer operator, runs it, and then afterwards performs greedy swaps of nodes to reduce the number of edge crossings further. While not perfect, it can improve the results of simpler heuristics.
• twolayerAgg - This aggregates the indices of ancestor nodes and orders nodes according to those aggregates. This is very fast and produces a good first-order decrossing, but is best when used in conjunction with twolayerGreedy.
• twolayerOpt - This is the sibling to decrossOpt that only optimizes a single layer, but otherwise has similar options. Just like full optimal decrossing, this will fail on large graphs, but also rarely produces better results than the combination of twolayerGreedy and twolayerAgg.

(default: twolayerGreedy)

get the current Twolayer for ordering

sets the number of passes to make

More passes may take longer, but might result in a better output.

(default: 24)

get the current number of passes