louvain_partitions generator issue

[Michael]

Hi all

I cannot explain the behavior of the following code. Why do I get different results when I iterate with a for loop the generator directly and when I iterate over the list of the generator?

----------------------------

import networkx as nx
import networkx.algorithms.community as nx_comm

G = nx.path_graph(200)

def evaluate(part):
    print(len(part))
    count = 0
    for nodes in part:
        if len(nodes) == 0:
            count += 1
    print(f"\tEmpty Clusters: {count}")

louvian_clustering = list(nx_comm.louvain_partitions(G, resolution=1, seed=20))
for part in louvian_clustering:
    evaluate(part)

print("----")

for part in nx_comm.louvain_partitions(G, resolution=1, seed=20):
    evaluate(part)

----------------------------

I get the following output:

----------------------------

90

    Empty Clusters: 76 

41

    Empty Clusters: 27 

17 

    Empty Clusters: 3

 14

    Empty Clusters: 0

 ----

 90

    Empty Clusters: 0

 41 

    Empty Clusters: 0 

17

    Empty Clusters: 0 

14 

    Empty Clusters: 0

----------------------------

Thanks for an explanation.

Michael

[Dan Schult]

Something is wrong here... perhaps to do with yield and return being used in the same function.

Can you open an issue at https://github.com/networkx/networkx/issues?

Thanks!