[Python] Efficiently converting edgelist to distance matrix in Python

R user learning Python here. I have an undirected unweighted edgelist and I want a distance matrix, showing the shortest path between all my nodes. There are about 30000 nodes and 40000 edges, and my goal is a 30000x30000 matrix where entry ij shows distance between node i and node j.

On my i7 16GB RAM PC I can achieve the result in 59 seconds in R 4.3.2

library(igraph)
g <- graph_from_data_frame(el, directed = F)
dm <- distances(g)


But neither attempt 1 nor 2 below finished after 10 minutes of running in Python 3.12.4 (Spyder 5.5.1):

import networkx as nx
import pandas as pd

g = nx.from_pandas_edgelist(el, 'mid', 'cid')

#Attempt 1
dm = nx.floyd_warshall_numpy(g)

#Attempt 2
gdict= dict(nx.all_pairs_shortest_path_length(g))
dm = pd.Dataframe(gdict)


Where el is a pandas Dataframe and mid and cid are uint16 columns.

Am I doing something wrong here, or is Python just very slow in shortest path algorithms compared to R? Alternatively, how can I create a distance matrix from an edgelist in Python efficiently?

My data looks something like this:

data = {
'mid': [1, 1, 2, 2, 3, 4],
'cid': [2, 3, 3, 4, 5, 5]
}
el = pd.DataFrame(data)

Continue reading...