[BUG] Layouts fail for directed graphs

Expected Behavior

Layouts should work with directed graphs (IMO) Perhaps not a bug in the sense that I’m not sure we promise this works for directed, but still. If we decided to only support undirected maybe a clearer error is warranted.

Actual Behavior

NetworkXNotImplemented: not implemented for directed type

Example Code

import networkx as nx
from graspologic.layouts import layout_umap

g = nx.erdos_renyi_graph(10, 0.7, directed=True)
layout_umap(g)

Full Traceback

---------------------------------------------------------------------------
NetworkXNotImplemented                    Traceback (most recent call last)
~/JHU_code/maggot/maggot_connectome/scripts/what_is_flow_rank.py in 
      3 
      4 g = nx.erdos_renyi_graph(10, 0.7, directed=True)
----> 5 layout_umap(g)

~/miniconda3/envs/maggot-revamp/lib/python3.8/site-packages/graspologic/layouts/auto.py in layout_umap(graph, min_dist, n_neighbors, max_edges, random_seed)
    173     """
    174 
--> 175     lcc_graph, tensors, labels = _node2vec_for_layout(graph, max_edges, random_seed)
    176     points = umap.UMAP(
    177         min_dist=min_dist, n_neighbors=n_neighbors, random_state=random_seed

~/miniconda3/envs/maggot-revamp/lib/python3.8/site-packages/graspologic/layouts/auto.py in _node2vec_for_layout(graph, max_edges, random_seed)
    214 ) -> Tuple[nx.Graph, np.ndarray, np.ndarray]:
    215     graph = _approximate_prune(graph, max_edges)
--> 216     graph = _largest_connected_component(graph)
    217 
    218     start = time.time()

~/miniconda3/envs/maggot-revamp/lib/python3.8/site-packages/graspologic/layouts/auto.py in _largest_connected_component(graph)
    182 
    183 def _largest_connected_component(graph: nx.Graph) -> nx.Graph:
--> 184     largest_component = max(nx.connected_components(graph), key=len)
    185     return graph.subgraph(largest_component).copy()
    186 

 in connected_components(G)

~/miniconda3/envs/maggot-revamp/lib/python3.8/site-packages/networkx/utils/decorators.py in _not_implemented_for(not_implement_for_func, *args, **kwargs)
     74         if match:
     75             msg = f"not implemented for {' '.join(graph_types)} type"
---> 76             raise nx.NetworkXNotImplemented(msg)
     77         else:
     78             return not_implement_for_func(*args, **kwargs)

NetworkXNotImplemented: not implemented for directed type

Your Environment

• Python version: 3.8
• graspologic version: dev

Additional Details

183 def _largest_connected_component(graph: nx.Graph) -> nx.Graph:
--> 184     largest_component = max(nx.connected_components(graph), key=len)

Just from these 2 lines on the traceback, I suspect this just needs to be adapted to check for weakly connected components. Or, use graspologic.utils.is_fully_connected from our own codebase. However in #112 we are talking about replacing the networkx-based check with a faster one from scipy. Not sure if the layouts code will always assume networkx input going forward. AFAIK there is nothing that inherently needs to be undirected for the layouts process, or am I missing something?

[bdpedigo]

FYI @kareef928 has been working on the conversion stuff so we’ll keep him in the loop too!

[daxpryce]

Would be interesting to see if the time-to-convert an nx graph into a csr matrix and run that lcc check is faster than doing it all in networkx. I’ll keep you guys posted after I crunch some numbers 😃