Merge branch 'empty_twitter_graph' into 'master'

Empty graph handling

See merge request imi-sad/spikexplore!1

Author: naspert

.gitlab-ci.yml

@@ -0,0 +1,29 @@
+image: python:3.9
+
+stages:
+  - test
+
+variables:
+    PIP_CACHE_DIR: "$CI_PROJECT_DIR/.cache/pip"
+
+cache:
+    paths:
+      - .cache/pip
+      - venv/
+
+before_script:
+  - export LANG=en_US.UTF-8
+  - export LANGUAGE=en_US:en
+  - export LC_ALL=en_US.UTF-8
+  - python -m pip install --upgrade pip
+  - python -m pip install virtualenv
+  - virtualenv venv
+  - source venv/bin/activate
+  - pip install --upgrade -r requirements.txt
+  - pip install black pytest
+
+Tests:
+  stage: test
+  script:
+    - python -m pytest tests
+

setup.py

@@ -2,7 +2,7 @@
 
 setup(
     name='spikexplore',
-    version='0.0.11',
+    version='0.0.12',
     description='Graph exploration using inhomogeneous filtered diffusion',
     url='https://github.com/epfl-lts2/spikexplore',
     author='Nicolas Aspert, Benjamin Ricaud',

spikexplore/collect_edges.py

@@ -144,11 +144,9 @@ def spiky_ball(initial_node_list, graph_handle, cfg,
         total_node_list = total_node_list + new_node_list
         edges_df_in, edges_df_out = split_edges(edges_df, total_node_list)
 
-        # Equivalent of add to graph
-        total_edges_df = total_edges_df.append(edges_df_in)
-        total_nodes_df = total_nodes_df.append(nodes_df)
-        # add the edges linking the new nodes
-        total_edges_df = total_edges_df.append(new_edges)
+        # add edges linking to new nodes
+        total_edges_df = pd.concat([total_edges_df, edges_df_in, new_edges])
+        total_nodes_df = pd.concat([total_nodes_df, nodes_df])
 
         new_node_list, new_edges = random_subset(edges_df_out, expansion_type, mode=random_subset_mode,
                                                  mode_value=random_subset_size, coeff=degree)
@@ -157,7 +155,8 @@ def spiky_ball(initial_node_list, graph_handle, cfg,
         logger.debug('new edges:{} subset:{} in_edges:{}'.format(len(edges_df_out), len(new_edges), len(edges_df_in)))
 
     logger.debug('Nb of layers reached: {}'.format(depth))
-    total_edges_df = total_edges_df.sort_values('weight', ascending=False)
+    if not total_edges_df.empty:
+        total_edges_df = total_edges_df.sort_values('weight', ascending=False)
 
     return total_node_list, total_nodes_df, total_edges_df, node_acc
 

spikexplore/graph.py

@@ -29,6 +29,8 @@ def convert_to_json(edge_df):
 def graph_from_edgeslist(edge_df, min_weight=0):
     logger.debug('Creating the graph from the edge list')
     # The indices in the dataframe are source and target for the edges
+    if edge_df.empty:
+        return nx.empty_graph()
     G = nx.from_pandas_edgelist(edge_df[edge_df['weight'] >= min_weight],
                                 source='source', target='target', create_using=nx.DiGraph)
     logger.info('Nb of nodes: {}'.format(G.number_of_nodes()))
@@ -133,10 +135,10 @@ def process_hop(graph_handle, node_list, nodes_info_acc):
         node_info, edges_df = graph_handle.get_neighbors(node)
         node_info, edges_df = graph_handle.filter(node_info, edges_df)
 
-        total_nodes_df = total_nodes_df.append(node_info.get_nodes())
+        total_nodes_df = pd.concat([total_nodes_df, node_info.get_nodes()])
         nodes_info_acc.update(node_info)  # add new info
 
-        total_edges_df = total_edges_df.append(edges_df)
+        total_edges_df = pd.concat([total_edges_df, edges_df])
         neighbors_dic = graph_handle.neighbors_with_weights(edges_df)
         new_node_dic = combine_dicts(new_node_dic, neighbors_dic)
 

spikexplore/graph_explore.py

@@ -7,6 +7,8 @@
 def create_graph(backend, nodes_df, edges_df, nodes_info, config):
     min_weight = config.min_weight
     g = graph_from_edgeslist(edges_df, min_weight=min_weight)
+    if nx.is_empty(g):
+        return g
     g = backend.add_graph_attributes(g, nodes_df, edges_df, nodes_info)
     g = reduce_graph(g, config.min_degree)
     g = handle_spikyball_neighbors(g, backend)

tests/twitter_test.py

@@ -43,4 +43,7 @@ def test_sampling_v2_coreball(self):
         self.assertTrue(g_sub.number_of_nodes() > 5)
         self.assertTrue(g_sub.number_of_edges() > 10)
 
-
+    def test_empty_graph(self):
+        g_sub, _ = graph_explore.explore(self.sampling_backend, ['#InvalidUsername'], self.sampling_config)
+        self.assertTrue(g_sub.number_of_nodes() == 0)
+        self.assertTrue(g_sub.number_of_edges() == 0) 

tests/wikipedia_test.py

@@ -46,3 +46,8 @@ def test_sampling_coreball(self):
         self.assertTrue(g_sub.number_of_edges() > 700)
         self.assertTrue(nx.is_connected(g_sub))
         self.assertTrue(set(g_sub.nodes()).intersection(self.wiki_config.pages_ignored) == set())
+
+    def test_empty_graph(self):
+        g_sub, _ = graph_explore.explore(self.sampling_backend, ['Non existent page of wikipedia forever'], self.sampling_config)
+        self.assertTrue(g_sub.number_of_nodes() == 0)
+        self.assertTrue(g_sub.number_of_edges() == 0)