def test_depth_0(self, msgsteiner):
'''
Run Forest with depth = 0; msgsteiner fails and test_util.run_forest raises an error
INPUT:
msgsteiner - fixture object with the value of --msgpath parsed by conftest.py
'''
params = copy.deepcopy(conf_params)
params['D'] = 0
with pytest.raises(Exception) as excinfo:
test_util.run_forest(msgsteiner, params, forest_opts)
assert 'Forest did not generate the optimal forest file' in excinfo.value.message
def test_w_missing(self, msgsteiner):
''' Run Forest with w missing from the configuration file and check
that Forest does not produce an output subnetwork
INPUT:
msgsteiner - fixture object with the value of --msgpath parsed by conftest.py
'''
params = copy.deepcopy(conf_params)
with pytest.raises(Exception) as excinfo:
test_util.run_forest(msgsteiner, params, forest_opts)
# Forest will not raise an Exception, it simply exits if w is missing
# Have to check for missing output files to determine whether it exited
assert 'Forest did not generate the optimal forest file' in excinfo.value.message
def test_w_1(self, msgsteiner):
''' Run Forest with w=1 and check optimal subnetwork
INPUT:
msgsteiner - fixture object with the value of --msgpath parsed by conftest.py
'''
params = copy.deepcopy(conf_params)
params['w'] = 1
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
# Check that the DiGraph has the expected properties
# Undirected edges are loaded as a pair of directed edges
assert graph.order() == 3, "Unexpected number of nodes"
assert graph.size() == 3, "Unexpected number of edges"
# Check that the DiGraph has the expected edges
assert graph.has_edge('A','C')
assert graph.has_edge('C','A')
assert graph.has_edge('C','D')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 2.0
# Excluded prizes: 0
# Edge costs: 1.0
# Number of trees * w: 1 * 1.0 = 1.0
assert isclose(2.0, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
def test_beta_2(self, msgsteiner):
'''
See test_beta_1; beta = 2 is enough to overcome the hub penalty so that the hub at A is chosen:
p'(A) = 2*5 - 2*4 = 2
'''
params = copy.deepcopy(conf_params)
params['b'] = 2
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
try:
assert graph.order() == 5, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
assert graph.has_edge('A', 'B')
assert graph.has_edge('A', 'C')
assert graph.has_edge('A', 'D')
assert graph.has_edge('A', 'E')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 1.4
# Excluded prizes: 0
# Edge costs: 0.4
# Number of trees * w: 1 * 1.0 = 1.0
assert isclose(1.4, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_beta_1(self, msgsteiner):
'''
In p'(v) = beta * p(v) - mu * deg(v), beta = 1 is too small to overcome the hub penalty with mu = 2:
p'(A) = 1*5 - 2*4 = -3
We expect forest to use the more costly edges BC and CD in its network instead
'''
params = copy.deepcopy(conf_params)
params['b'] = 1
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
try:
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 2, "Unexpected number of edges"
assert graph.has_edge('B', 'C')
assert graph.has_edge('D', 'E')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 0.8
# Excluded prizes: -3.0
# Edge costs: 1.8
# Number of trees * w: 2 * 1.0 = 2.0
assert isclose(0.8, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_beta_026(self, msgsteiner):
'''
Run Forest with beta = 0.26
See test_beta_024; this value of beta makes it worth while to obtain prizes for C and D;
we expect the following network:
A B
\ \
C D
note an undirected edge ab in a digraph is represented by the network containing both edges ab and ba
'''
params = copy.deepcopy(conf_params)
params['b'] = 0.26
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 4
assert graph.size() == 4
assert graph.has_edge('A', 'C')
assert graph.has_edge('C', 'A')
assert graph.has_edge('B', 'D')
assert graph.has_edge('D', 'B')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 0.5
# Excluded prizes: 0
# Edge costs: 0.5
# Number of trees * w: 0 * 2 = 0
assert isclose(0.5, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
def test_beta_1(self, msgsteiner):
'''
In p'(v) = beta * p(v) - mu * deg(v), beta = 1 is too small to overcome the hub penalty with mu = 2:
p'(A) = 1*5 - 2*4 = -3
We expect forest to use the more costly edges BC and CD in its network instead
'''
params = copy.deepcopy(conf_params)
params['b'] = 1
graph, objective = test_util.run_forest(msgsteiner, params,
forest_opts)
try:
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 2, "Unexpected number of edges"
assert graph.has_edge('B', 'C')
assert graph.has_edge('D', 'E')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 0.8
# Excluded prizes: -3.0
# Edge costs: 1.8
# Number of trees * w: 2 * 1.0 = 2.0
assert isclose(0.8, objective, rtol=0,
atol=1e-5), 'Incorrect objective function value'
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_beta_2(self, msgsteiner):
'''
See test_beta_1; beta = 2 is enough to overcome the hub penalty so that the hub at A is chosen:
p'(A) = 2*5 - 2*4 = 2
'''
params = copy.deepcopy(conf_params)
params['b'] = 2
graph, objective = test_util.run_forest(msgsteiner, params,
forest_opts)
try:
assert graph.order() == 5, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
assert graph.has_edge('A', 'B')
assert graph.has_edge('A', 'C')
assert graph.has_edge('A', 'D')
assert graph.has_edge('A', 'E')
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 1.4
# Excluded prizes: 0
# Edge costs: 0.4
# Number of trees * w: 1 * 1.0 = 1.0
assert isclose(1.4, objective, rtol=0,
atol=1e-5), 'Incorrect objective function value'
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_depth_3_w_1(self, msgsteiner):
'''
When the solver is constrained by depth, it still picks up net positive nodes.
This results in a different network than if the solver is not constrained by depth.
Note behavior of the following tests for comparison (which we do not repeat in this class):
TestW#test_w_1, which has w = 1 and D = 5
we are not constrained by depth and we pick A - C -> D
TestDepth#test_depth_10, which has w = 0 and D = 10
there is no forest penalty, so the optimal network is the same as this one
'''
params = copy.deepcopy(conf_params)
params['D'] = 3
params['w'] = 1
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
# Check that the DiGraph has the expected properties
# Undirected edges are loaded as a pair of directed edges
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
# Check that the DiGraph has the expected edges
assert graph.has_edge('A','C')
assert graph.has_edge('C','A')
assert graph.has_edge('B','D')
assert graph.has_edge('D','B')
def test_depth_3_w_1(self, msgsteiner):
'''
When the solver is constrained by depth, it still picks up net positive nodes.
This results in a different network than if the solver is not constrained by depth.
Note behavior of the following tests for comparison (which we do not repeat in this class):
TestW#test_w_1, which has w = 1 and D = 5
we are not constrained by depth and we pick A - C -> D
TestDepth#test_depth_10, which has w = 0 and D = 10
there is no forest penalty, so the optimal network is the same as this one
'''
params = copy.deepcopy(conf_params)
params['D'] = 3
params['w'] = 1
graph, objective = test_util.run_forest(msgsteiner, params,
forest_opts)
# Check that the DiGraph has the expected properties
# Undirected edges are loaded as a pair of directed edges
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
# Check that the DiGraph has the expected edges
assert graph.has_edge('A', 'C')
assert graph.has_edge('C', 'A')
assert graph.has_edge('B', 'D')
assert graph.has_edge('D', 'B')
def test_depth_2(self, msgsteiner):
'''
Run Forest with depth = 2; while this depth is valid, it results in no edges
because the depth parameter is used as a strict inequality and there is a dummy node
'''
params = copy.deepcopy(conf_params)
params['D'] = 2
graph, objective = test_util.run_forest(msgsteiner, params,
forest_opts)
assert graph.order() == 0, "Unexpected number of nodes"
assert graph.size() == 0, "Unexpected number of edges"
def test_beta_024(self, msgsteiner):
'''
Run Forest with beta = 0.24
This value of beta is just below the 0.25 value of beta which sets the prize benefit
of C and D equal to the cost of obtaining those prizes through edges AC and BD: we expect
an empty network again
'''
params = copy.deepcopy(conf_params)
params['b'] = 0.24
graph = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 0, "Unexpected number of nodes"
assert graph.size() == 0, "Unexpected number of edges"
def test_beta_0(self, msgsteiner):
'''
Run Forest with beta = 0 and check optimal subnetwork
In p'(v) = beta * p(v) - mu * deg(v), beta = 0 implies that an empty graph is the optimal network
INPUT:
msgsteiner - fixture object with the value of --msgpath parsed by conftest.py
'''
params = copy.deepcopy(conf_params)
params['b'] = 0
graph = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 0, "Unexpected number of nodes"
assert graph.size() == 0, "Unexpected number of edges"
def test_depth_10(self, msgsteiner):
'''
Run forest with depth = 10; the results are the same as in test_depth_3
'''
params = copy.deepcopy(conf_params)
params['D'] = 3
graph, objective = test_util.run_forest(msgsteiner, params,
forest_opts)
# Check that the DiGraph has the expected properties
# Undirected edges are loaded as a pair of directed edges
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
# Check that the DiGraph has the expected edges
assert graph.has_edge('A', 'C')
assert graph.has_edge('C', 'A')
assert graph.has_edge('B', 'D')
assert graph.has_edge('D', 'B')
def test_beta_2(self, msgsteiner):
'''
See test_beta_1; beta = 2 is enough to overcome the hub penalty so that the hub at A is chosen:
p'(A) = 2*5 - 2*4 = 2
'''
params = copy.deepcopy(conf_params)
params['b'] = 2
graph = test_util.run_forest(msgsteiner, params, forest_opts)
try:
assert graph.order() == 5, "Unexpected number of nodes"
assert graph.size() == 4, "Unexpected number of edges"
assert graph.has_edge('A', 'B')
assert graph.has_edge('A', 'C')
assert graph.has_edge('A', 'D')
assert graph.has_edge('A', 'E')
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_beta_1(self, msgsteiner):
'''
In p'(v) = beta * p(v) - mu * deg(v), beta = 1 is too small to overcome the hub penalty with mu = 2:
p'(A) = 1*5 - 2*4 = -3
We expect forest to use the more costly edges BC and CD in its network instead
'''
params = copy.deepcopy(conf_params)
params['b'] = 1
graph = test_util.run_forest(msgsteiner, params, forest_opts)
try:
assert graph.order() == 4, "Unexpected number of nodes"
assert graph.size() == 2, "Unexpected number of edges"
assert graph.has_edge('B', 'C')
assert graph.has_edge('D', 'E')
except AssertionError as e:
test_util.print_graph(graph)
raise e
def test_beta_024(self, msgsteiner):
'''
Run Forest with beta = 0.24
This value of beta is just below the 0.25 value of beta which sets the prize benefit
of C and D equal to the cost of obtaining those prizes through edges AC and BD: we expect
an empty network again
'''
params = copy.deepcopy(conf_params)
params['b'] = 0.24
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 0, "Unexpected number of nodes"
assert graph.size() == 0, "Unexpected number of edges"
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 0.48
# Excluded prizes: 0.48
# Edge costs: 0
# Number of trees * w: 0
assert isclose(0.48, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
def test_beta_0(self, msgsteiner):
'''
Run Forest with beta = 0 and check optimal subnetwork
In p'(v) = beta * p(v) - mu * deg(v), beta = 0 implies that an empty graph is the optimal network
INPUT:
msgsteiner - fixture object with the value of --msgpath parsed by conftest.py
'''
params = copy.deepcopy(conf_params)
params['b'] = 0
graph, objective = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 0, "Unexpected number of nodes"
assert graph.size() == 0, "Unexpected number of edges"
# Check that the optimal forest has the correct objective function
# value, using isclose to allow for minor floating point variation
# Objective function: 0
# Excluded prizes: 0
# Edge costs: 0
# Number of trees * w: 0
assert isclose(0, objective, rtol=0, atol=1e-5), 'Incorrect objective function value'
def test_beta_026(self, msgsteiner):
'''
Run Forest with beta = 0.26
See test_beta_024; this value of beta makes it worth while to obtain prizes for C and D;
we expect the following network:
A B
\ \
C D
note an undirected edge ab in a digraph is represented by the network containing both edges ab and ba
'''
params = copy.deepcopy(conf_params)
params['b'] = 0.26
graph = test_util.run_forest(msgsteiner, params, forest_opts)
assert graph.order() == 4
assert graph.size() == 4
assert graph.has_edge('A', 'C')
assert graph.has_edge('C', 'A')
assert graph.has_edge('B', 'D')
assert graph.has_edge('D', 'B')
