What kind of dataset can use the louvain of graphscope?

[neblen]

I use the example dataset of graphscope and use the project method for getting a subgraph->
simple_graph = g.project(vertices={"person": []}, edges={"knows": ["dist"]})
but it can't work with the louvain of graphscope.
1.code is following
import graphscope
graphscope.set_option(show_log=True)
graphscope.set_option(log_level="DEBUG")
sess = graphscope.session(addr='x.x.x.x:xx', k8s_service_type="LoadBalancer")
g = sess.g()
g = g.add_vertices("/tmp/testingdata/property/p2p-31_property_v_0", label="person")
g = g.add_edges("/tmp/testingdata/property/p2p-31_property_e_0", label="knows")
print(g.schema)
simple_graph = g.project(vertices={"person": []}, edges={"knows": ["dist"]})
print(simple_graph)
from graphscope import louvain
result_context=louvain(simple_graph)
2.erro is following

InvalidArgumentError Traceback (most recent call last)
in
1 from graphscope import louvain
----> 2 result_context=louvain(simple_graph)

in louvain(graph, min_progress, progress_tries)

/opt/conda/lib/python3.8/site-packages/graphscope/framework/app.py in project_to_simple(func, *args, **kwargs)
46 if graph.graph_type == types_pb2.ARROW_PROPERTY:
47 graph = graph._project_to_simple()
---> 48 return func(graph, *args[1:], **kwargs)
49
50

in louvain(graph, min_progress, progress_tries)

/opt/conda/lib/python3.8/site-packages/graphscope/framework/app.py in _not_compatible_for(not_compatible_for_func, *args, **kwargs)
98 )
99 else:
--> 100 return not_compatible_for_func(*args, **kwargs)
101
102 return _not_compatible_for

/opt/conda/lib/python3.8/site-packages/graphscope/analytical/app/louvain.py in louvain(graph, min_progress, progress_tries)
65 """
66 if graph.is_directed():
---> 67 raise InvalidArgumentError("Louvain not support directed graph.")
68 return AppAssets(algo="louvain")(graph, min_progress, progress_tries)

InvalidArgumentError: 'Louvain not support directed graph.'

[sighingnow]

I think it is an expected error. Moved to discussion.

[sighingnow]

@siyuan0322 Any further comment/doc on louvain algorithm?

[acezen]

the louvain algorithm now only support undirected graph, you can use louvain like:

g = sess.g(directed=False)
g = g.add_vertices("/tmp/testingdata/property/p2p-31_property_v_0", label="person")
g = g.add_edges("/tmp/testingdata/property/p2p-31_property_e_0", label="knows")
simple_graph = g.project(vertices={"person": []}, edges={"knows": ["dist"]})
from graphscope import louvain
result_context=louvain(simple_graph)

[neblen]

thank you so much
this example can work in notebook from ali's jupyter notebook.
If person want to work in personal cluster ,you need to give your notebook more resources.