Python NetworkX: Confining force-directed layout within circular boundary

Question

Python NetworkX has a method spring_layout that simulates a force-directed representation of a NetworkX instance; however, this leads to an adjusted network with nodes that are not confined within a particular boundary shape (e.g., a circle). Below is an example of this (notice how the overall graph shape is arbitrary, albeit smaller clusters are visible):

Is it possible to set a boundary limit such that any node seeking to position itself beyond the set boundary is confined to the boundary edge? Below is an example illustrating this (notice how nodes along the perimeter seem to lie against an invisible circular border):

Answer 1

Maybe you can convert the Cartesian coordinates (x, y) to Polar coordinates (t, r) from pos returned by nx.spring_layout:

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

# G is your graph
G = nx.balanced_tree(10, 3)
pos = nx.spring_layout(G)

# Extract Cartesian coordinates
x, y = np.vstack(list(pos.values())).T

# Convert to Polar coordinates
r = np.sqrt(x**2, y**2)
t = np.arctan2(y, x)

# Create a figure with polar projection
fig, ax = plt.subplots(subplot_kw={'projection': 'polar'})

# Remove some annoying elements
ax.grid(False)
ax.set_xticklabels([])
ax.set_yticklabels([])

# Rebuild the position dictionary
pos2 = dict(zip(pos.keys(), [np.array(c) for c in zip(t, r)]))

# Draw graph
nx.draw_networkx(G, pos2, ax=ax, with_labels=False)

Polar coordinates:

Cartesian coordinates: