def test_random_directed_1(self):
CC = nx.from_numpy_matrix(self.C, create_using=nx.DiGraph)
lenC = len(max(nx.kosaraju_strongly_connected_components(CC), key=len))
DD = nx.from_numpy_matrix(self.D, create_using=nx.DiGraph)
lenD = len(max(nx.kosaraju_strongly_connected_components(DD), key=len))
self.assertEqual(lenC, (self.C).shape[0])
self.assertEqual(lenD, (self.D).shape[0])
def DONEnolongerneeded():
"""extracting a part of the metabolic network
dataset from Stanford University
output: a connected graph (contains loops, I verifyed)
of 932 nodes, saved into json file. The BASE.
"""
f_ll = open("PP-Pathways_ppi.csv", "rt").readlines()
allEdges = set()
for line in f_ll:
tup = tuple(int(i) for i in line.replace("\n", "").split(","))
if tup[0] > 2500 or tup[1] > 2500:
continue
else:
allEdges.add(tup)
ALLMETAB = nx.DiGraph()
ALLMETAB.add_edges_from(list(allEdges))
print(len(list((ALLMETAB.edges()))))
print(len(list((ALLMETAB.nodes()))))
#extract the nodes making part of the largest connected component:
ho_l = max(nx.kosaraju_strongly_connected_components(ALLMETAB), key=len)
print(type(ho_l))
SG = ALLMETAB.subgraph(ho_l)
print(len(SG.nodes()))
print(len(SG.edges()))
print(nx.is_connected(SG.to_undirected())) #True, ok
print(nx.is_directed(SG)) # , True ok
EDGEStxt = ", ".join(list(str(list(tu)) for tu in SG.edges()))
textjson = '{"nodes": ' + str(len(
SG.nodes())) + ', "edges": [' + EDGEStxt + ']}'
with open("METABO.jsonlike", "w") as p:
p.write(textjson)
p.close()
return "done"
def test_null_graph(self):
G = nx.DiGraph()
assert_equal(list(nx.strongly_connected_components(G)), [])
assert_equal(list(nx.kosaraju_strongly_connected_components(G)), [])
assert_equal(list(nx.strongly_connected_components_recursive(G)), [])
assert_equal(len(nx.condensation(G)), 0)
assert_raises(nx.NetworkXPointlessConcept, nx.is_strongly_connected, nx.DiGraph())
def test_null_graph(self):
G = nx.DiGraph()
assert list(nx.strongly_connected_components(G)) == []
assert list(nx.kosaraju_strongly_connected_components(G)) == []
assert list(nx.strongly_connected_components_recursive(G)) == []
assert len(nx.condensation(G)) == 0
pytest.raises(nx.NetworkXPointlessConcept, nx.is_strongly_connected, nx.DiGraph())
def test_algo_time_nx(self):
"""
this test check the run time of connected_components and shortest path
in networkx
:return:None
"""
ga1 = GraphAlgo.GraphAlgo()
filename = '../data/G_100_800_1.json'
ga1.load_from_json(filename)
ga = self.graph_nx(ga1.get_graph())
with self.elapsed_timer() as elapsed:
star = elapsed()
l = nx.single_source_dijkstra(ga, 0, 2)
end = elapsed()
res = end - star
try:
with open("res2.txt", "a") as f:
f.write(f"\ngraph:{filename},shortest_path:{res}\n{l}")
star = elapsed()
# print(list(nx.strongly_connected_components(ga)))
print(list(nx.kosaraju_strongly_connected_components(ga)))
end = elapsed()
res = end - star
f.write(f"\ngraph:{filename},connected_component:{res}\n")
except IOError as e:
print(e)
def find_SCCS(self):
"""
(1) It first finds the SCCs of the transaction graph.
verified
"""
# scc requires a digraph
graph = self.create_graph(_type='di')
self.SCCs = list(kosaraju_strongly_connected_components(graph))
self.SCCs.sort()
def city_graph_generator(city):
"""
Generate an OSMnx graph of **city**.
Graph modified to be undirected, weighted, and meeting the triangle
inequality.
Parameters
----------
city: str
The city to query road network from. *'City, Country'*.
Returns
-------
G: graph
OSMnx generated graph of **city**.
Graph is weighted, undirected, and satisfies the triangle inequality.
Raises
------
AssertionError
If **city** is not a string.
"""
# Assert city input is string.
assert type(city) is str, "City name not valid, must be string."
# Query and generate OSMNX graph from city.
O = ox.graph_from_place(city,
network_type='drive',
simplify=True,
retain_all=False,
timeout=60)
# Keep only the largest strongly connected component.
keep = max(nx.kosaraju_strongly_connected_components(O), key=len)
G_sub = O.subgraph(keep).copy()
G_dir = nx.DiGraph(G_sub)
G = nx.Graph()
G.graph.update(G_dir.graph)
# Iterate over edges to keep maximum weight edges
# in the case where there are parallel edges.
for u, v in G_dir.edges():
G.add_nodes_from([(u, G_dir.nodes[u]), (v, G_dir.nodes[v])])
if (v, u) in G_dir.edges():
wgt = max(G_dir.edges[u, v]['length'], G_dir.edges[v, u]['length'])
else:
wgt = G_dir.edges[u, v]['length']
G.add_edges_from([(u, v, G_dir.edges[u, v])])
G.edges[u, v]['length'] = wgt
G.edges[u, v]['weight'] = wgt
return G
def stronglyConnectedComponents(self) -> list:
"""
Gets strongly connected components of graph.
Returns
-------
list
Strongly connected component of specified node
"""
return list(nodeSet for nodeSet in
nx.kosaraju_strongly_connected_components(self.graph))
def test_connected_raise(self):
G = nx.Graph()
with pytest.raises(NetworkXNotImplemented):
next(nx.strongly_connected_components(G))
with pytest.raises(NetworkXNotImplemented):
next(nx.kosaraju_strongly_connected_components(G))
with pytest.raises(NetworkXNotImplemented):
next(nx.strongly_connected_components_recursive(G))
pytest.raises(NetworkXNotImplemented, nx.is_strongly_connected, G)
pytest.raises(nx.NetworkXPointlessConcept, nx.is_strongly_connected,
nx.DiGraph())
pytest.raises(NetworkXNotImplemented, nx.condensation, G)
def build(self):
"""Finalise the graph, after adding all input files to it."""
assert not self.final, 'Trying to mutate a final graph.'
# Replace each strongly connected component with a single node `NodeSet`
for scc in sorted(nx.kosaraju_strongly_connected_components(self.graph),
key=len, reverse=True):
if len(scc) == 1:
break
self.shrink_to_node(NodeSet(scc))
self.final = True
def createGraphs2(subdata, deg):
avg = 0
for j in range(len(subdata)):
matrix = subdata[j]
threshold = alzTest.findThresh(matrix, deg)
for k in range(len(subdata[j])):
for l in range(len(subdata[j][k])):
if subdata[j][k][l] > threshold:
subdata[j][k][l] = 1
else:
subdata[j][k][l] = 0
G = nx.DiGraph(subdata[j])
listoflists = nx.kosaraju_strongly_connected_components(G)
avg += len(listoflists[0])
return float(avg / len(subdata))
def createGraphs2(subdata, deg):
avg = 0
for j in range(len(subdata)):
matrix = subdata[j]
threshold = alzTest.findThresh(matrix, deg)
for k in range(len(subdata[j])):
for l in range(len(subdata[j][k])):
if subdata[j][k][l] > threshold:
subdata[j][k][l] = 1
else:
subdata[j][k][l] = 0
G = nx.DiGraph(subdata[j])
listoflists = nx.kosaraju_strongly_connected_components(G)
avg += len(listoflists[0])
return float(avg/len(subdata))
def print_results(nx_g, cm_g):
""" Print out result by each algorithm.
For testing/debuging purposes only.
"""
print('Tarjan class:', [sorted(component) for component in tarjan.Tarjan(cm_g).get_scc()])
print()
print('Tarjan func: ', [sorted(component) for component in tarjan.strongly_connected_components(cm_g)])
print()
print('Gabow class:', [sorted(component) for component in gabow.Gabow(cm_g).get_scc()])
print()
print('Gabow func: ', [sorted(component) for component in gabow.strongly_connected_components(cm_g)])
print()
print('Tarjan NX: ', [sorted(component) for component in nx.strongly_connected_components(nx_g)])
print()
print('Kosaraju NX:', [sorted(component) for component in nx.kosaraju_strongly_connected_components(nx_g)])
def degvssizeAvg(data):
degreedict = {}
for deg in range(1, 100, 5):
subdata = copy.deepcopy(data[('NL', 'corr')])
threshold = determineThreshold(subdata, deg)
print "degree: " + str(deg) + " threshold: " + str(threshold)
myList = createGraphs(subdata, threshold)
avg = 0
for graph in myList:
listoflists = nx.kosaraju_strongly_connected_components(graph)
avg += len(listoflists[0])
avg = float(avg/len(myList))
degreedict[deg] = avg
plt.plot(degreedict.keys(), degreedict.values(), 'bo')
plt.show()
def degvssizeAvg(data):
degreedict = {}
for deg in range(1, 100, 5):
subdata = copy.deepcopy(data[('NL', 'corr')])
threshold = determineThreshold(subdata, deg)
print "degree: " + str(deg) + " threshold: " + str(threshold)
myList = createGraphs(subdata, threshold)
avg = 0
for graph in myList:
listoflists = nx.kosaraju_strongly_connected_components(graph)
avg += len(listoflists[0])
avg = float(avg / len(myList))
degreedict[deg] = avg
plt.plot(degreedict.keys(), degreedict.values(), 'bo')
plt.show()
def stronglyConnectedSubgraphs(self) -> list[Graph[V, E]]:
"""
Gets strongly connected subgraphs of graph.
Returns
-------
list[Graph[V, E]]
Graph strongly connected components
"""
subgraphs: list[Graph[V, E]] = list()
for nodeSet in nx.kosaraju_strongly_connected_components(self.graph):
subgraph: Graph[V, E] = Graph(is_directed=self.isDirected,
is_weighted=self.isWeighted)
subgraph.addNodes(nodeSet)
for edge in self.graph.edges:
if edge[0] in nodeSet and edge[1] in nodeSet:
subgraph.addEdge(edge[0], edge[1])
subgraphs.append(subgraph)
return subgraphs
def directed(self):
D = self.Initial_Laplacian(self.size, self.prob)
for i in range(self.size):
if D[i][i] == 1.:
D[i][i] = 0.
""" Until here, we have the adjacency matrix of arandom directed graph,
but we are not sure whether it is stringly connected or not! """
graph = nx.from_numpy_matrix(D, create_using=nx.DiGraph)
largest = max(nx.kosaraju_strongly_connected_components(graph), key=len)
adj = np.zeros((len(largest), len(largest)))
v = 0
w = 0
for i in largest:
for j in largest:
adj[v][w] = D[i][j]
w +=1
w = 0
v +=1
return adj
def topological_order_value_iteration(P, states, actions, gamma, eps, max_iteration):
'''
A function that performs topological value iteration on the MDP input.
INPUT:
- P the transition table
P[s][a] == [(probability, nextstate, reward, done), ...]
- states, the discrete sets of states in the environment
- actions, the discrete set of actions in the environment
- gamma, the discount factor for rewards
OUTPUT:
value function values for discrete states
'''
g, P = build_graph(P, actions, states)
V = defaultdict(int)
# returned in reversed topological order
comps = list(nx.kosaraju_strongly_connected_components(g))
comps.reverse()
for comp in comps:
print("component is: {}".format(comp))
# update the VI values when you add each component
V.update(hacky_value_iteration(P, comp, actions, V, gamma, eps, max_iteration))
return V
#!/usr/bin/env python2 import networkx as nx import graph, tarjan, gabow if __name__ == '__main__': for i in range(3): # generate random graph and convert to custom representation nx_graph = nx.gnm_random_graph(15, 25, directed=True) custom_graph = graph.nx2custom(nx_graph) # get strong connected components print "Strongly connected components by Tarjan's algorithm:" print [ sorted(component) for component in tarjan.strongly_connected_components(custom_graph) ] print print "Strongly connected components by Gabow's algorithm:" print [ sorted(component) for component in gabow.strongly_connected_components(custom_graph) ] print print "Strongly connected components by Tarjan's networkx library algorithm:" print [ sorted(component) for component in nx.strongly_connected_components(nx_graph) ] print print "Strongly connected components by Kosaraju's networkx library algorithm:" print [ sorted(component) for component in nx.kosaraju_strongly_connected_components(nx_graph) ] print '-----------------------------------------------------------------------'
def test_Expo_directed_2(self):
C = Exponential(10).directed()
CC = nx.from_numpy_matrix(C, create_using=nx.DiGraph)
lenC = len(max(nx.kosaraju_strongly_connected_components(CC), key=len))
self.assertEqual(lenC, C.shape[0])
print("nx: ", f)
start = time()
algo.shortest_path(1, 2)
print(f'shortest_path-Python: {time() - start}')
print("algo: ", algo.shortest_path(1, 2))
"""---------------connected components compare---------------"""
print("test 1-SCC")
g = compare_nx.load_from_json('../data/G_10_80_0.json')
algo = GraphAlgo()
algo.load_from_json('../data/G_10_80_0.json')
start = time()
f = nx.kosaraju_strongly_connected_components(g)
print(f'scc-nx: {time() - start}')
start = time()
algo.connected_components()
print(f'scc-Python: {time() - start}')
print("test 2-SCC")
g = compare_nx.load_from_json('../data/G_100_800_0.json')
algo = GraphAlgo()
algo.load_from_json('../data/G_100_800_0.json')
start = time()
f = nx.kosaraju_strongly_connected_components(g)
print(f'scc-nx: {time() - start}')
def run_simulation():
nx_results = [
] # contains the result (in sec) of all the functions in networkx (Short|Component|Components)
py_results = [
] # contains the result (in sec) of all the functions in python (Short|Component|Components)
# contains the result (in sec) of all the functions in java (Short|Component|Components)
java_results = [
0.0024046, 0.0020393, 0.0286995, 0.3990609, 0.5003517, 0.2987889,
0.0001906, 0.000615, 0.0005584, 0.0255793, 0.0330737, 0.0009482,
0.0001255, 0.00016747, 0.0255405, 0.2244528, 0.757112, 0.9485512
]
"""-------------------- Shortest_path comparison --------------------"""
json_graph = [
'../data/G_10_80_1.json', '../data/G_100_800_1.json',
'../data/G_1000_8000_1.json', '../data/G_10000_80000_1.json',
'../data/G_20000_160000_1.json', '../data/G_30000_240000_1.json'
]
i = 1
for f in json_graph:
print('----- test', i, ': Shortest_path comparison -----')
i = i + 1
graph_nx = compare_nx.load_from_json(f)
graph_algo = GraphAlgo()
graph_algo.load_from_json(f)
time_start = time.time()
src = int(uniform(0, graph_algo.graph.v_size()))
dest = int(uniform(0, graph_algo.graph.v_size()))
list_nx = nx.shortest_path(graph_nx, src, dest, weight='weight')
time_end = time.time()
delta = time_end - time_start
nx_results.append(delta)
print('Time_nx:', delta)
print('graph_nx: ', list_nx)
time_start = time.time()
graph_algo.shortest_path(src, dest)
time_end = time.time()
delta = time_end - time_start
py_results.append(delta)
print('Time_python:', delta)
print('graph_algo:', graph_algo.shortest_path(src, dest), '\n')
"""-------------------- Connected_component comparison --------------------"""
i = 1
for f in json_graph:
print('----- test', i, ': Connected_component (Id_scc) -----')
i = i + 1
graph_algo = GraphAlgo()
graph_algo.load_from_json(f)
time_start = time.time()
random = int(uniform(0, graph_algo.graph.v_size()))
component_list = graph_algo.connected_component(random)
time_end = time.time()
delta = time_end - time_start
py_results.append(delta)
# print("component_list:", component_list)
print('Time_python:', delta, '\n')
nx_results.append(
0) # Connected_component does not exist in nx -> always zero
"""-------------------- Connected_components comparison --------------------"""
i = 1
for f in json_graph:
print('----- test', i, ': Connected_componentS (ALL_scc) -----')
i = i + 1
graph_nx = compare_nx.load_from_json(f)
graph_algo = GraphAlgo()
graph_algo.load_from_json(f)
time_start = time.time()
list_nx = nx.kosaraju_strongly_connected_components(graph_nx)
time_end = time.time()
delta = time_end - time_start
nx_results.append(delta)
print('Time_nx:', delta)
time_start = time.time()
graph_algo.connected_components()
time_end = time.time()
delta = time_end - time_start
py_results.append(delta)
print('Time_python:', delta, '\n')
# print('----- results -----')
# print("nx_results:", nx_results)
# print("py_results:", py_results)
# print("java_results:", java_results)
return nx_results, py_results, java_results
def graph_check(graph_matrix, size):
graph = nx.from_numpy_matrix(graph_matrix, create_using=nx.DiGraph)
largest = max(nx.kosaraju_strongly_connected_components(graph), key=len)
assert size == len(largest)
def compute_SCC_graphs(p):
graph = nx.DiGraph()
t_triples, t_cardinality = hdt.search_triples("", p, "")
print("amount of triples: ", t_cardinality)
for (l, p, r) in t_triples:
graph.add_edge(l, r)
# compute SCC
start = time.time()
mydict = {}
for n in graph.nodes:
collect_succssor = []
for s in graph.successors(n):
collect_succssor.append(s)
mydict[n] = collect_succssor
sccs = tarjan(mydict)
filter_scc = [x for x in sccs if len(x) > 1]
# compute the counter for ths filtered SCC
ct = Counter()
for f in filter_scc:
ct[len(f)] += 1
print('SCC info by tarjan : ', ct)
end = time.time()
hours, rem = divmod(end - start, 3600)
minutes, seconds = divmod(rem, 60)
print("Time taken Tarjan: {:0>2}:{:0>2}:{:05.2f}".format(
int(hours), int(minutes), seconds))
start = time.time()
nx_sccs = nx.strongly_connected_components(graph)
filter_nx_sccs = [x for x in nx_sccs if len(x) > 1]
nx_ct = Counter()
for f in filter_nx_sccs:
nx_ct[len(f)] += 1
print('SCC info by original nx : ', nx_ct)
end = time.time()
hours, rem = divmod(end - start, 3600)
minutes, seconds = divmod(rem, 60)
print("Time taken original: {:0>2}:{:0>2}:{:05.2f}".format(
int(hours), int(minutes), seconds))
start = time.time()
nx_sccs = nx.kosaraju_strongly_connected_components(graph)
filter_nx_sccs = [x for x in nx_sccs if len(x) > 1]
nx_ct = Counter()
for f in filter_nx_sccs:
nx_ct[len(f)] += 1
print('SCC info by original nx : ', nx_ct)
end = time.time()
hours, rem = divmod(end - start, 3600)
minutes, seconds = divmod(rem, 60)
print("Time taken kosaraju: {:0>2}:{:0>2}:{:05.2f}".format(
int(hours), int(minutes), seconds))
# obtain the corresponding scc graphs
scc_graphs = []
for s in filter_scc:
g = graph.subgraph(s).copy()
scc_graphs.append(g)
return filter_scc, scc_graphs
