Tips for writing my own layering

[kelemeno]

Tips for writing my own layering

[kelemeno]

Thanks this is great!

I will use the depthLayering method you provided, I have the height stored. (I don't think the first one will work, as I will not draw every edge of the tree, only a section of the tree. But I do want to keep the tree's layering. I also don't think it is compact).

And fortunately not a multigraph.

[erikbrinkman]

@kelemeno sorry for the late response!

• If your tree is compact, e.g. all children are one height below their parents, then simplex should just do the right thing, and so should longestPath (the later being a little faster).
• If it's not compact, you should be able to get away with just specifying ranks. It's still more computationally expensive than necessary, but it's the next easiest for your effort.
• The fastest for large trees will be writing it yourself. If you look at the documentation for LayeringOperators it has a pretty simple requirement

  After calling a layering operator on a dag, every node's value should be set to a non-negative integer layer, starting at 0. Additionally children should have a strictly larger layer than their parents, and if a node has multiple links to the same child, it's layer should be at least two greater than its parent.
  If you had the depth stored in the node data you could use that with something like:

  function depthLayering(dag: Dag<{ myDepth: number }, unknown>): void {
      for (const node of dag) {
          node.value = node.data.myDepth;
      }
  }

  However, if you don't have the height stored, and want it purely in graph terms, then you should also be able to use height because that does exactly what you want.

  function depthLayering(dag: Dag<unknown, unknown>): void {
      dag.height();
  }

  Do note, that both of these implementation assume that you don't have a muultigraph, e.g. you can't have two edges between the same two nodes. Due to the way we layout multi-edges, we have to insert dummy layers, and these implementations don't do that.