Create graph with curved and labelled edges in Networkx

Question

I'm currently displaying directed graphs with few nodes and edges connecting them, with nx.draw. The edges are labelled via nx.draw_networkx_edge_labels.

Now I wanted to "lighten" the "rigidity" aspect of the graph by setting the connectionstyle, which works fine with non-labelled edges.

Problem is that if I display labels, they are drawn as if the edge was not curved, which ends up creating a huge offset between the edge and the label.

Is there any way to work around this limitation? I could not find an "offset" option to nx.draw_networkx_edge_labels to address this issue.

EDIT:

Above is a quick example of the issue:

import matplotlib.pyplot as plt
import networkx as nx

tab = ("r", ["s", "t", "u", "v", "w", "x", "y", "z"])

producer = tab[0]
consumers = tab[1]

color_map = []
DG = nx.DiGraph()
for i, cons in enumerate(consumers):
    DG.add_edge(producer, cons, label=f"edge-{i}")

for i in range(len(DG.nodes())):
    if i < 1 + len(consumers):
        color_map.append("#DCE46F")
    else:
        color_map.append("#6FA2E4")
pos = nx.shell_layout(DG)
labels = nx.get_edge_attributes(DG, 'label')
nx.draw(DG, pos, node_color=color_map, connectionstyle="arc3, rad=0.2", with_labels=True, font_size=8, node_size=1000, node_shape='o')
nx.draw_networkx_edge_labels(DG, pos, edge_labels=labels)

plt.show()

current output:

Answer 1

If you are open to using other libraries for the visualization, I wrote (and maintain) netgraph. In netgraph, edge labels track edges even if they are curved.

import numpy as np
import matplotlib.pyplot as plt
import networkx as nx

from netgraph import Graph # pip install netgraph

tab = ("r", ["s", "t", "u", "v", "w", "x", "y", "z"])

producer = tab[0]
consumers = tab[1]

DG = nx.DiGraph()
for i, cons in enumerate(consumers):
    DG.add_edge(producer, cons, label=f"edge-{i}")

node_color = dict()
for node in DG:
    if node in producer:
        node_color[node] = "#DCE46F"
    else:
        node_color[node] = "#6FA2E4"

pos = nx.shell_layout(DG)
pos[producer] = pos[producer] + np.array([0.2, 0])
edge_labels = nx.get_edge_attributes(DG, 'label')

Graph(DG, node_layout=pos, edge_layout='curved', origin=(-1, -1), scale=(2, 2),
      node_color=node_color, node_size=8.,
      node_labels=True, node_label_fontdict=dict(size=10),
      edge_labels=edge_labels, edge_label_fontdict=dict(size=10),
)
plt.show()