Okay, so, if I understand correctly, the purpose of that condition would be to check whether all nodes have an attribute named _igraph_index in the NetworkX graph, right? I don't think that next(iter(g.nodes)) (as suggested in this PR) would do the trick as g.nodes seems to iterate over the node names.

I think the correct fix would be next(iter(nxg.nodes.values())). This is a bit convoluted, but basically nxg.nodes returns a dict-like view that maps NetworkX node names to their "data dictionaries"; next(iter(...)) then takes the first element of it, which is the data dictionary of the first node.

yep... shall we squash & merge?