def test_hash(self):
d1 = Domain("d", "foo", [1, 2, 3])
h1 = hash(d1)
self.assertIsNotNone(h1)
d2 = Domain("d", "foo", [1, 2, 4])
h2 = hash(d2)
self.assertIsNotNone(h2)
self.assertNotEqual(h1, h2)
d3 = Domain("d", "foo", [1, 2, 3])
h3 = hash(d3)
self.assertEqual(h1, h3)
def test_api_dcop_graph_coloring():
# Graph coloring with 3 variables and color preferences
dcop = DCOP('test')
# Domain and variables
d = Domain('color', '', ['R', 'G'])
variables = create_variables('v', [1, 2, 3], d)
# Creating constraints using the convenience += notation
# We could also use the more verbose dcop.add_constraint method.
# Variable(s) and domain(s) involved in the constraint are automatically
# added to the dcop.
# unary constraints for preferences
dcop += 'cost_1', '-0.1 if v1 == "R" else 0.1 ', variables
dcop += 'cost_2', '-0.1 if v2 == "G" else 0.1 ', variables
dcop += 'cost_3', '-0.1 if v3 == "G" else 0.1 ', variables
# coloring constraints : v1 != v2 != v3
dcop += 'c1', '1 if v1 == v2 else 0', variables
dcop += 'c2', '1 if v3 == v2 else 0', variables
# let's check the dcop really has all required domain, variables and
# constraints
assert len(dcop.variables) == 3
assert 'v1' in dcop.variables
assert len(dcop.domains) == 1
assert 'color' in dcop.domains
assert len(dcop.constraints) == 5
assert 'cost_3' in dcop.constraints
assert 'c2' in dcop.constraints
def test_create_several_variables_from_list(self):
d = Domain('color', '', ['R', 'G', 'B'])
vars = create_variables('x_', ['a1', 'a2', 'a3'], d)
self.assertIn('x_a1', vars)
self.assertTrue(isinstance(vars['x_a2'], Variable))
self.assertEqual(vars['x_a3'].name, 'x_a3')
def test_comp_creation_with_factory_method():
d = Domain("d", "", ["R", "G"])
v1 = Variable("v1", d)
v2 = Variable("v2", d)
c1 = constraint_from_str("c1", "10 if v1 == v2 else 0", [v1, v2])
graph = build_computation_graph(None, constraints=[c1], variables=[v1, v2])
comp_node = graph.computation("c1")
algo_def = AlgorithmDef.build_with_default_param("maxsum")
comp_def = ComputationDef(comp_node, algo_def)
comp = build_computation(comp_def)
assert comp is not None
assert comp.name == "c1"
assert comp.factor == c1
comp_node = graph.computation("v1")
algo_def = AlgorithmDef.build_with_default_param("maxsum")
comp_def = ComputationDef(comp_node, algo_def)
comp = build_computation(comp_def)
assert comp is not None
assert comp.name == "v1"
assert comp.variable.name == "v1"
assert comp.factors == ["c1"]
def test_create_several_variables_from_range(self):
d = Domain("color", "", ["R", "G", "B"])
variables = create_variables("x_", range(10), d)
self.assertIn("x_1", variables)
self.assertTrue(isinstance(variables["x_2"], Variable))
self.assertEqual(variables["x_3"].name, "x_3")
def test_create_several_variables_from_list(self):
d = Domain("color", "", ["R", "G", "B"])
variables = create_variables("x_", ["a1", "a2", "a3"], d)
self.assertIn("x_a1", variables)
self.assertTrue(isinstance(variables["x_a2"], Variable))
self.assertEqual(variables["x_a3"].name, "x_a3")
def test_create_several_variables_from_several_lists(self):
d = Domain('color', '', ['R', 'G', 'B'])
vars = create_variables('m_', (['x1', 'x2'], ['a1', 'a2', 'a3']), d)
self.assertEqual(len(vars), 6)
self.assertIn(('x1', 'a2'), vars)
self.assertTrue(isinstance(vars[('x2', 'a3')], Variable))
self.assertEqual(vars[('x2', 'a3')].name, 'm_x2_a3')
def test_build_graph_from_variables_constraints():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
v2 = Variable('v2', d)
v3 = Variable('v3', d)
c1 = constraint_from_str('c1', 'v1 * 0.5 + v2 - v3', [v1, v2, v3])
graph = build_computation_graph(variables=[v1, v2, v3], constraints=[c1])
def test_create_several_variables_from_several_lists(self):
d = Domain("color", "", ["R", "G", "B"])
variables = create_variables("m_", (["x1", "x2"], ["a1", "a2", "a3"]), d)
self.assertEqual(len(variables), 6)
self.assertIn(("x1", "a2"), variables)
self.assertTrue(isinstance(variables[("x2", "a3")], Variable))
self.assertEqual(variables[("x2", "a3")].name, "m_x2_a3")
def test_create_ok(self):
d1 = Domain('d1', '', [1, 2, 3, 5])
v1 = Variable('v1', d1)
f1 = constraint_from_str('f1', 'v1 * 0.5', [v1])
cv1 = VariableComputationNode(v1, [f1])
cf1 = FactorComputationNode(f1)
cg = ComputationsFactorGraph([cv1], [cf1])
def generate_ising(
row_count: int,
col_count: int,
bin_range: float,
un_range: float,
extensive: bool,
no_agents: bool,
fg_dist: bool,
var_dist: bool,
) -> Tuple[DCOP, Dict, Dict]:
grid_graph = nx.grid_2d_graph(row_count, col_count, periodic=True)
domain = Domain("var_domain", "binary", [0, 1])
variables = generate_binary_variables(grid_graph, domain)
constraints = {}
unary_constraints = generate_unary_constraints(variables, un_range, extensive)
constraints.update(unary_constraints)
binary_constraints = generate_binary_constraints(
grid_graph, variables, bin_range, extensive
)
constraints.update(binary_constraints)
agents = {}
fg_mapping = defaultdict(lambda: [])
var_mapping = defaultdict(lambda: [])
for (row, col) in grid_graph.nodes:
agent = AgentDef(f"a_{row}_{col}")
agents[agent.name] = agent
left = (row - 1) % row_count
down = (col + 1) % col_count
if var_dist:
var_mapping[agent.name].append(f"v_{row}_{col}")
if fg_dist:
fg_mapping[agent.name].append(f"v_{row}_{col}")
fg_mapping[agent.name].append(f"cu_v_{row}_{col}")
# Sort coordinate to make sure we build the name in the same order as when
# creating the constraints:
(r1, c1), (r2, c2) = sorted([(row, col), (left, col)])
fg_mapping[agent.name].append(f"cb_v_{r1}_{c1}_v_{r2}_{c2}")
(r1, c1), (r2, c2) = sorted([(row, col), (row, down)])
fg_mapping[agent.name].append(f"cb_v_{r1}_{c1}_v_{r2}_{c2}")
name = f"Ising_{row_count}_{col_count}_{bin_range}_{un_range}"
if no_agents:
agents = {}
dcop = DCOP(
name,
domains={"var_domain": domain},
variables={v.name: v for v in variables.values()},
agents=agents,
constraints=constraints,
)
return dcop, dict(var_mapping), dict(fg_mapping)
def test_simple_repr(self):
d = Domain("d", "foo", [1, 2, 3])
r = simple_repr(d)
print(r)
self.assertEqual(r["__qualname__"], "Domain")
self.assertEqual(r["__module__"], "pydcop.dcop.objects")
self.assertEqual(r["name"], "d")
self.assertEqual(r["domain_type"], "foo")
def test_raise_when_duplicate_computation_name(self):
d1 = Domain('d1', '', [1, 2, 3, 5])
v1 = Variable('v1', d1)
# here we create a relation with the same name as the variable
f1 = constraint_from_str('v1', 'v1 * 0.5', [v1])
cv1 = VariableComputationNode(v1, ['f1'])
cf1 = FactorComputationNode(f1)
self.assertRaises(KeyError, ComputationsFactorGraph, [cv1], [cf1])
def test_create_node_with_custom_name():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
c1 = constraint_from_str('c1', 'v1 * 0.5', [v1])
n1 = VariableComputationNode(v1, [c1], name='foo')
assert v1 == n1.variable
assert n1.name == 'foo'
def generate_powerlaw_var_constraints(
num_var: int, domain_size: int, constraint_range: int
) -> Tuple[Dict[str, Variable], Dict[str, Constraint], Domain]:
"""
Generate variables and constraints for a power-law based constraints
graph.
All constraints are binary and the graph is generated using the Barabasi
Albert method.
Parameters
----------
num_var: int
number of variables
domain_size: int
size of the domain of the variables
constraint_range: int
range in which constraints take their value (uniform random value of
ech possible assignment).
Returns
-------
A tuple with variables, constraints and domain.
"""
# Use a barabasi powerlaw based constraints graph
graph = nx.barabasi_albert_graph(num_var, 2)
# import matplotlib.pyplot as plt
# plt.subplot(121)
# nx.draw(graph) # default spring_layout
# plt.show()
domain = Domain("d", "d", range(domain_size))
variables = {}
for n in graph.nodes:
v = Variable(var_name(n), domain)
variables[v.name] = v
logger.debug("Create var for node %s : %s", n, v)
constraints = {}
for i, (n1, n2) in enumerate(graph.edges):
v1 = variables[var_name(n1)]
v2 = variables[var_name(n2)]
values = random_assignment_matrix([v1, v2], range(constraint_range))
c = NAryMatrixRelation([v1, v2], values, name=c_name(n1, n2))
logger.debug("Create constraints for edge (%s, %s) : %s", v1, v2, c)
constraints[c.name] = c
logger.info(
"Generates %s variables and %s constraints in a powerlaw" "network",
len(variables),
len(constraints),
)
return variables, constraints, domain
def test_simple_repr(self):
d = Domain('d', 'foo', [1, 2, 3])
r = simple_repr(d)
print(r)
self.assertEqual(r['__qualname__'], 'Domain')
self.assertEqual(r['__module__'], 'pydcop.dcop.objects')
self.assertEqual(r['name'], 'd')
self.assertEqual(r['domain_type'], 'foo')
self.assertEqual(r['values'], [1, 2, 3])
def test_compute_factor_cost_at_start():
d = Domain("d", "", ["R", "G"])
v1 = Variable("v1", d)
v2 = Variable("v2", d)
c1 = constraint_from_str("c1", "10 if v1 == v2 else 0", [v1, v2])
obtained = factor_costs_for_var(c1, v1, {}, "min")
assert obtained["R"] == 0
assert obtained["G"] == 0
assert len(obtained) == 2
def test_one_var_one_factor():
dcop = DCOP('test', 'min')
d1 = Domain('d1', '--', [1, 2, 3])
v1 = Variable('v1', d1)
dcop += 'c1', '0.5 * v1', [v1]
g = build_computation_graph(dcop)
assert len(g.links) == 1
assert len(g.nodes) == 2
def test_density_two_var_one_factor():
dcop = DCOP('test', 'min')
d1 = Domain('d1', '--', [1, 2, 3])
v1 = Variable('v1', d1)
v2 = Variable('v2', d1)
dcop += 'c1', '0.5 * v1 + v2', [v1, v2]
g = build_computation_graph(dcop)
assert g.density() == 4 / 6
def test_create_node_no_neigbors():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
c1 = constraint_from_str('c1', 'v1 * 0.5', [v1])
n1 = VariableComputationNode(v1, [c1])
assert v1 == n1.variable
assert c1 in n1.constraints
assert len(n1.links) == 1 # link to our-self
assert not n1.neighbors
def test_var_node_simple_repr():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
c1 = constraint_from_str('c1', 'v1 * 0.5', [v1])
cv1 = VariableComputationNode(v1, [c1])
r = simple_repr(cv1)
cv1_obtained = from_repr(r)
assert cv1 == cv1_obtained
def test_get_candidate_value_selected():
d = Domain("d", "vals", ["vB", "vD", "vA", "vE"])
v = Variable("v", d)
obtained = get_value_candidates(v, "vB")
assert obtained == ["vD", "vA", "vE"]
obtained = get_value_candidates(v, "vA")
assert obtained == ["vE"]
obtained = get_value_candidates(v, "vE")
assert obtained == []
def test_create_node_with_binary_constraint():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
v2 = Variable('v2', d)
c1 = constraint_from_str('c1', 'v1 * 0.5 - v2', [v1, v2])
n1 = VariableComputationNode(v1, [c1])
assert v1 == n1.variable
assert c1 in n1.constraints
assert list(n1.links)[0].has_node('v2')
assert 'v2' in n1.neighbors
def test_variablenode_simple_repr():
d1 = Domain('d1', '', [1, 2, 3, 5])
v1 = Variable('v1', d1)
f1 = constraint_from_str('f1', 'v1 * 0.5', [v1])
cv1 = VariableComputationNode(v1, ['f1'])
cf1 = FactorComputationNode(f1, )
r = simple_repr(cv1)
obtained = from_repr(r)
assert obtained == cv1
assert cv1.variable == obtained.variable
def test_select_value_no_cost_var():
d = Domain("d", "", ["R", "G", "B"])
v1 = Variable("v1", d)
selected, cost = select_value(v1, {}, "min")
assert selected in {"R", "G", "B"}
assert cost == 0
v1 = VariableWithCostFunc("v1", [1, 2, 3], lambda v: (4 - v) / 10)
selected, cost = select_value(v1, {}, "min")
assert selected == 3
assert cost == 0.1
def create_dcop():
dcop = DCOP('test')
# Domain and vraibales
d = Domain('color', '', ['R', 'G'])
variables = create_variables('v', [1, 2, 3], d)
# unary constraints for preferences
dcop += 'cost_1', '-0.1 if v1 == "R" else 0.1 ', variables
dcop += 'cost_2', '-0.1 if v2 == "G" else 0.1 ', variables
dcop += 'cost_3', '-0.1 if v3 == "G" else 0.1 ', variables
# coloring constraints : v1 != v2 != v3
dcop += 'c1', '1 if v1 == v2 else 0', variables
dcop += 'c2', '1 if v3 == v2 else 0', variables
return dcop
def test_graph_density():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
v2 = Variable('v2', d)
v3 = Variable('v3', d)
c1 = constraint_from_str('c1', 'v1 * 0.5 + v2 - v3', [v1, v2, v3])
dcop = DCOP('dcop_test', 'min')
dcop.add_constraint(c1)
graph = build_computation_graph(dcop)
density = graph.density()
assert density == 1 / 3
def setUp(self):
variables = list(
create_variables('v', ['1', '2', '3'], Domain('d', '',
[1, 2])).values())
all_diff = constraint_from_str('all_diff', 'v1 + v2 + v3 ', variables)
v1, v2, v3 = variables
c1 = VariableComputationNode(v1, [all_diff])
c2 = VariableComputationNode(v2, [all_diff])
c3 = VariableComputationNode(v3, [all_diff])
nodes = [c1, c2, c3]
# links = [ConstraintLink('all_diff', ['c1', 'c2', 'c3'])]
self.cg = ComputationConstraintsHyperGraph(nodes)
self.agents = [AgentDef('a1'), AgentDef('a2'), AgentDef('a3')]
def generate_secp(args):
logger.info("Generate SECP %s", args)
light_count = args.lights
model_count = args.models
rule_count = args.rules
capacity = args.capacity
max_model_size = args.max_model_size
max_rule_size = args.max_rule_size
light_domain = Domain("light", "light", range(0, 5))
lights_var, lights_cost = build_lights(light_count, light_domain)
models_var, models_constraints = build_models(
light_domain, lights_var, max_model_size, model_count
)
rules_constraints = build_rules(rule_count, lights_var, models_var, max_rule_size)
# Agents : one for each light
agents = build_agents(lights_var, lights_cost, capacity)
# Force
# * each light variable to be hosted on the corresponding agent
# * model constraint and var are preferred on the same agent
variables = lights_var.copy()
variables.update(models_var)
constraints = models_constraints.copy()
constraints.update(lights_cost)
constraints.update(rules_constraints)
dcop = DCOP(
"graph coloring",
"min",
domains={"light_domain": light_domain},
variables=variables,
agents=agents,
constraints=constraints,
)
if args.output:
outputfile = args.output
write_in_file(outputfile, dcop_yaml(dcop))
else:
print(dcop_yaml(dcop))
def test_link_simple_repr():
d = Domain('d', 'test', [1, 2, 3])
v1 = Variable('v1', d)
v2 = Variable('v2', d)
v3 = Variable('v3', d)
c1 = constraint_from_str('c1', 'v1 * 0.5 + v2 - v3', [v1, v2, v3])
cv1 = VariableComputationNode(v1, [c1])
cv2 = VariableComputationNode(v2, [c1])
cv3 = VariableComputationNode(v3, [c1])
link = ConstraintLink(c1.name, ['c1', 'c2', 'c3'])
r = simple_repr(link)
link_obtained = from_repr(r)
assert link == link_obtained
