LayeredLayout.NetworkSimplex

type NEdge n = { delta :: Int, eid :: Int, src :: n, tgt :: n, weight :: Number }

An edge for the network simplex algorithm.

• delta – the minimum span (target.layer − source.layer ≥ delta).
• weight – contribution to the optimisation objective. Network simplex minimises the weighted sum of weight * (layer span) across all edges, so heavier edges are kept shorter.
• eid – a per-edge integer index, unique within the input array; the algorithm uses it as a key into cutvalue / visited sets.

runNetworkSimplex :: forall n. Ord n => Array n -> Array (NEdge n) -> Map n Int

Run the network simplex over the given nodes and edges.

The caller is responsible for ensuring the input is weakly- connected (split into components, or inject an artificial root that dominates all sources). When the input is disconnected the algorithm still terminates, but the layering of the unreached component(s) is not optimised.

fuzzyStZero :: Number

Tolerance for floating-point imprecision when checking whether a cut value is negative. Port of ELK's FUZZY_ST_ZERO = -1e-10.