"Membership cannot exceed number of nodes" error with leidenalg 0.8.4

[AvantiShri]

Hi, I'm getting an odd error with the latest version of leidenalg that wasn't appearing with 0.8.3. Please see this github gist for a reproduction of the error:

https://colab.research.google.com/gist/AvantiShri/46060b5fe0fac540830a3078758a487b/-membership-cannot-exceed-number-of-nodes-error-with-leidenalg-0-8-4.ipynb

Consider the following test:

import numpy as np
import leidenalg
import igraph as ig

def test():
  g = ig.Graph(directed=None)
  g.add_vertices(2)
  g.add_edges([(0,1)])
  g.es['weight'] = np.array([0.5])
  leidenalg.find_partition(graph=g, partition_type=leidenalg.ModularityVertexPartition,
      weights=g.es['weight'], n_iterations=-1,
      initial_membership=np.array([0,1]), seed=1234)

With leidenalg 0.8.3, test() runs without any error. However, with version 0.8.4, it produces:

/usr/local/lib/python3.7/dist-packages/leidenalg/functions.py in find_partition(graph, partition_type, initial_membership, weights, n_iterations, max_comm_size, seed, **kwargs)
     86   partition = partition_type(graph,
     87                              initial_membership=initial_membership,
---> 88                              **kwargs)
     89   optimiser = Optimiser()
     90 

/usr/local/lib/python3.7/dist-packages/leidenalg/VertexPartition.py in __init__(self, graph, initial_membership, weights, node_sizes)
    467 
    468     self._partition = _c_leiden._new_ModularityVertexPartition(pygraph_t,
--> 469         initial_membership, weights, node_sizes)
    470     self._update_internal_membership()
    471 

TypeError: Membership cannot exceed number of nodes.

[vtraag]

Hi @AvantiShri, thanks for reporting this! I now realise that I changed something in 0.8.4 to correct some warnings, but that was previously used to accommodate for numpy. I will correct this again. As a workaround, please pass in any arguments, including the initial_membership, function as a Python list, not as a numpy array. You can use .tolist() of the numpy array.

[AvantiShri]

For future people encountering this: @vtraag's suggestion of using .tolist() (rather than simply wrapping the numpy array in a list(...) call) seems to be important.