Can't display correct bipartite graph with networkx - networkx

Question

I am trying to draw a bipartite graph with nodes on left colored differently than nodes on right. I am using networkx and matplotlib to do so.
Given a bipartite graph [(1, 3), (2, 5), (3, 4)], I wish to display [1,2,3], colored blue on one side and [4,5] colored aqua on other side, with the edges (1, 3), (2, 5), (3, 4) in between.
The following is my code.
import networkx as nx
import matplotlib.pyplot as plt
def draw_bipartite(edges_list):
left,right = set(),set()
for s,t in edges_list:
right.add(s)
left.add(t)
B = nx.Graph()
B.add_nodes_from(list(right), bipartite=0)
B.add_nodes_from(list(left), bipartite=1)
B.add_edges_from(edges_list)
nodecolor = []
for node in B.nodes():
a = 'blue' if node in list(right) else 'aqua'
nodecolor.append(a)
l,r = nx.bipartite.sets(B)
pos = {}
pos.update((node, (1, index)) for index, node in enumerate(l))
pos.update((node, (2, index)) for index, node in enumerate(r))
nx.draw(B, pos=pos,with_labels = True,node_color=nodecolor)
plt.show()
draw_bipartite([(1, 3), (2, 5), (3, 4)])
In output, the group [1,2,3] does not remain on left side, how do I keep it left side, as well as colored blue?

Answer 1

update
Looking at your code again, I see that only the color is determined by what you call left and right. The positions are determined according to what you call l and r. You have a bug in how you handle left and right and in how you handle l and r, so I'm addressing each of these separately.
So let's look at the color first: when you define left and right there is an error conceptually.
You are putting node 3 on both the left and the right. Your edge (1,3) puts 3 on the left while (3,4) puts it on the right. So the color that node 3 ends up with is 'blue' because you've put it in right. Since 4 was not put in right, it ends up 'aqua'. Note, it's confusing that what's on the left of a tuple ends up in right and you tell it to color right with blue, while saying in your explanation of the error that you want the left of your plot to be blue. You should make this all consistent as it's going to be difficult to avoid making more errors in the future.
Now let's look at node positions: why [1,2,3] does not end up on the left. You've defined l and r rather than taking the left and right you defined earlier (I don't think this was a good idea - use just one definition, or else you're just making it harder to hunt down bugs). You've used l,r = nx.bipartite.sets(B) There are a few issues here. First, the nodes 2 and 5 could end up on either side. It's an arbitrary choice that is determined by the unpredictable order in which python loops through keys in a dict. So you got lucky that 2 ended up in l. Similarly, node 1 could have easily ended up in r.
The expectation that 3 would end up on the same side as 1 is hopeless - the networkx algorithm bipartite.sets separates nodes based on whether they share edges or not. Since the edge (1,3) is in your graph, they will not end up on the same side. If you had used your previously defined left and right, then 1 would have been on the right, and 3 would have ended up on either the left or the right, depending on the order of your two pos.update commands because it is in both left and right.