Working with family structures in networkx

Question

I'm trying to represent and traverse graphs in order to show and control family relationships. I've been able to build the edges to make the following graph (How to draw a tree more beautifully in networkx plus countless others have gotten me this far - my red numbers of course)

So I understand this to be a Directed Graph with Parent relationships shown. I want to be able to get some more information using networkx but I haven't been able to work out how to approach it. I'd even take some good doco/tutorials if people have them but I'm trying to solve the following:

1. I've been able to identify the children who are also parents, can I split these into seperate graphs? e.g. 2,3,6,7,8,9,10,11,12 are one "family", 5,9,15 are another "family", 1,5 is also a family

2. Can I color or otherwise show the generations. I see 3 generations here (1,2,3 and 4-12 and 13-15) I believe I can use a topological sort to find these orders, however when I do I get 4 nodes, then 8 then 3 instead of 3,9,3

Answer 1

W.r.t. 2): Topological sort will always put a node in the highest possible layer, which is why in your case node 4 is put into in the top layer (presumably). Instead, you can map your y value from the dot layout to a layer. If that is too hacky for you, you can compute the the hierarchical layout explicitly. Both approaches are outlined here.

W.r.t. 1) You seem to define a "family" purely based on parent-child relationships, so all you need to do is compute the predecessors for each node, and then aggregate. Something along the lines of:

families = dict() # maps parents to children
for child in G:
    parents = tuple([predecessor for predecessor, target in nx.predecessor(G, child, cutoff=1).items() if target != []])
    if parents in families:
        families[parents].append(child)
    else: 
        families[parents] = [child]