在没有像 Efficient Sugiyama 这样的绘图算法的情况下，以树的形式布置 DAG 中的顶点（即顶部没有入边的顶点、仅依赖于下一层的顶点等）相当简单。但是，是否有一种简单的算法可以最大限度地减少边缘交叉？（对于某些图，可能不可能完全消除边缘交叉。）一张图片说一千个单词，那么有没有一种算法可以建议一些不交叉边缘的东西。（与此相比）。
编辑：结果
我已经接受了 Senthil 的建议 graphviz/dot —— 快速浏览一下文档确认它很容易用作库或外部工具，并且输出格式非常容易解析。但是，我最终选择使用GraphSharp，因为我已经在使用 .NET 等（尽管它绝对没有 dot 强大）。结果“足够好”，并且可以通过一点边缘路由和调整来做得更好（模糊的文本是因为3.5 WPF）。
这是完整的C# 代码（这是所有引用 QuickGraph 或 GraphSharp 的代码——是的；就是这么简单）：
internal static class LayoutManager
{
private const string ALGORITHM_NAME = "EfficientSugiyama";
private const bool MINIMIZE_EDGE_LENGTH = true;
private const double VERTEX_DISTANCE = 25;
private const double LAYER_DISTANCE = 25;
private const double MIN_CANVAS_OFFSET = 20;
public static void doLayout(GraphCanvas canvas)
{
// TODO use a background thread
// TODO add comments
canvas.IsEnabled = false;
canvas.Cursor = Cursors.Wait;
var graph = new BidirectionalGraph<GraphNode, LayoutEdge>();
var positions = new Dictionary<GraphNode, Point>();
var sizes = new Dictionary<GraphNode, Size>();
foreach(var node in canvas.nodes)
{
var size = node.RenderSize;
graph.AddVertex(node);
positions.Add(node, new Point(node.left + size.Width / 2, node.top + size.Height / 2));
sizes.Add(node, size);
}
foreach(var edge in canvas.edges)
{
graph.AddEdge(new LayoutEdge(edge));
}
var context = new LayoutContext<GraphNode, LayoutEdge, BidirectionalGraph<GraphNode, LayoutEdge>>(graph, positions, sizes, LayoutMode.Simple);
var parameters = new EfficientSugiyamaLayoutParameters();
parameters.VertexDistance = VERTEX_DISTANCE;
parameters.MinimizeEdgeLength = MINIMIZE_EDGE_LENGTH;
parameters.LayerDistance = LAYER_DISTANCE;
var factory = new StandardLayoutAlgorithmFactory<GraphNode, LayoutEdge, BidirectionalGraph<GraphNode, LayoutEdge>>();
var algorithm = factory.CreateAlgorithm(ALGORITHM_NAME, context, parameters);
algorithm.Compute();
canvas.deselectAll();
var minx = algorithm.VertexPositions.Select(kvp => kvp.Value.X - (kvp.Key.RenderSize.Width / 2)).Aggregate(Math.Min);
var miny = algorithm.VertexPositions.Select(kvp => kvp.Value.Y - (kvp.Key.RenderSize.Height / 2)).Aggregate(Math.Min);
minx -= MIN_CANVAS_OFFSET;
miny -= MIN_CANVAS_OFFSET;
minx = minx < 0 ? -minx : 0;
miny = miny < 0 ? -miny : 0;
foreach(var kvp in algorithm.VertexPositions)
{
var node = kvp.Key;
var pos = kvp.Value;
node.left = (pos.X - (node.RenderSize.Width / 2)) + minx;
node.top = (pos.Y - (node.RenderSize.Height / 2)) + miny;
}
canvas.Cursor = Cursors.Arrow;
canvas.IsEnabled = true;
}
private sealed class LayoutEdge : IEdge<GraphNode>
{
private readonly ConnectingEdge _edge;
public LayoutEdge(ConnectingEdge edge) { _edge = edge; }
public GraphNode Source { get { return _edge.output.node; } }
public GraphNode Target { get { return _edge.input.node; } }
}