def test_nonnegativity_transformation_1(self):
self.model.a = Var()
self.model.b = Var(within=NonNegativeIntegers)
self.model.c = Var(within=Integers, bounds=(-2,3))
self.model.d = Var(within=Boolean)
self.model.e = Var(domain=Boolean)
instance=self.model.create_instance()
xfrm = TransformationFactory('core.nonnegative_vars')
transformed = xfrm.create_using(instance)
# Check that all variables have nonnegative bounds or domains
for c in ('a', 'b', 'c', 'd', 'e'):
var = transformed.__getattribute__(c)
for ndx in var:
self.assertTrue(self.nonnegativeBounds(var[ndx]))
# Check that discrete variables are still discrete, and continuous
# continuous
for ndx in transformed.a:
self.assertIs(transformed.a[ndx].domain, NonNegativeReals)
for ndx in transformed.b:
self.assertIs(transformed.b[ndx].domain, NonNegativeIntegers)
for ndx in transformed.c:
self.assertIs(transformed.c[ndx].domain, NonNegativeIntegers)
for ndx in transformed.d:
self.assertIs(transformed.d[ndx].domain, Binary)
for ndx in transformed.e:
self.assertIs(transformed.e[ndx].domain, Binary)
def test_relax_integrality_cloned(self):
# Coverage of the _clear_attribute method
self.model.A = RangeSet(1,4)
self.model.a = Var()
self.model.b = Var(within=self.model.A)
self.model.c = Var(within=NonNegativeIntegers)
self.model.d = Var(within=Integers, bounds=(-2,3))
self.model.e = Var(within=Boolean)
self.model.f = Var(domain=Boolean)
instance=self.model.create_instance()
instance_cloned = instance.clone()
xfrm = TransformationFactory('core.relax_integer_vars')
rinst = xfrm.create_using(instance_cloned)
self.assertEqual(type(rinst.a.domain), RealSet)
self.assertEqual(type(rinst.b.domain), RealSet)
self.assertEqual(type(rinst.c.domain), RealSet)
self.assertEqual(type(rinst.d.domain), RealSet)
self.assertEqual(type(rinst.e.domain), RealSet)
self.assertEqual(type(rinst.f.domain), RealSet)
self.assertEqual(rinst.a.bounds, instance_cloned.a.bounds)
self.assertEqual(rinst.b.bounds, instance_cloned.b.bounds)
self.assertEqual(rinst.c.bounds, instance_cloned.c.bounds)
self.assertEqual(rinst.d.bounds, instance_cloned.d.bounds)
self.assertEqual(rinst.e.bounds, instance_cloned.e.bounds)
self.assertEqual(rinst.f.bounds, instance_cloned.f.bounds)
def test_relax_integrality2(self):
# Coverage of the _clear_attribute method
self.model.A = RangeSet(1, 4)
self.model.a = Var([1, 2, 3], dense=True)
self.model.b = Var([1, 2, 3], within=self.model.A, dense=True)
self.model.c = Var([1, 2, 3], within=NonNegativeIntegers, dense=True)
self.model.d = Var([1, 2, 3],
within=Integers,
bounds=(-2, 3),
dense=True)
self.model.e = Var([1, 2, 3], within=Boolean, dense=True)
self.model.f = Var([1, 2, 3], domain=Boolean, dense=True)
instance = self.model.create_instance()
xfrm = TransformationFactory('core.relax_integer_vars')
rinst = xfrm.create_using(instance)
self.assertEqual(type(rinst.a[1].domain), RealSet)
self.assertEqual(type(rinst.b[1].domain), RealSet)
self.assertEqual(type(rinst.c[1].domain), RealSet)
self.assertEqual(type(rinst.d[1].domain), RealSet)
self.assertEqual(type(rinst.e[1].domain), RealSet)
self.assertEqual(type(rinst.f[1].domain), RealSet)
self.assertEqual(rinst.a[1].bounds, instance.a[1].bounds)
self.assertEqual(rinst.b[1].bounds, instance.b[1].bounds)
self.assertEqual(rinst.c[1].bounds, instance.c[1].bounds)
self.assertEqual(rinst.d[1].bounds, instance.d[1].bounds)
self.assertEqual(rinst.e[1].bounds, instance.e[1].bounds)
self.assertEqual(rinst.f[1].bounds, instance.f[1].bounds)
def run_bilevel(self, *_args, **kwds):
os.chdir(currdir)
try:
#
# Import the model file to create the model
#
usermodel = pyutilib.misc.import_file(_args[0], clear_cache=True)
instance = usermodel.model
#
# Collected fixed variables
#
_fixed = kwds.pop('fixed', [])
for v in _fixed:
v_ = ComponentUID(v).find_component_on(instance)
fixed = [
ComponentUID(v).find_component_on(instance) for v in _fixed
]
#
# Apply transformations
#
if 'transform' in kwds:
xfrm = TransformationFactory(kwds['transform'])
transform_kwds = kwds.get('transform_kwds', {})
transform_kwds['fixed'] = fixed
new_instance = xfrm.create_using(instance, **transform_kwds)
else:
new_instance = instance
if self.solve:
#
# Solve the problem
#
opt = SolverFactory(kwds.get('solver', 'glpk'))
results = opt.solve(new_instance)
new_instance.solutions.store_to(results)
with open('result.yml', 'w') as ofile:
results.write(ostream=ofile, format='json')
elif kwds.get('format', 'lp') == 'lp':
#
# Write the file
#
io_options = {}
io_options['symbolic_solver_labels'] = True
io_options['file_determinism'] = 2
new_instance.name = 'Test'
new_instance.write(self.problem + "_result.lp",
io_options=io_options)
else:
#
# This is a hack. When we cannot write a valid LP file, we still
# write with the LP suffix to simplify the testing logic.
#
with open(self.problem + "_result.lp", 'w') as ofile:
new_instance.pprint(ostream=ofile)
except:
print("Failed to construct and transform the model")
raise
def test_relax_integrality_simple_cloned(self):
self.model.x = Var(within=Integers, bounds=(-2,3))
instance = self.model.create_instance()
instance_cloned = instance.clone()
xfrm = TransformationFactory('core.relax_discrete')
rinst = xfrm.create_using(instance_cloned)
self.assertIs(rinst.x.domain, Reals)
self.assertEqual(rinst.x.bounds, (-2,3))
self.assertIs(instance.x.domain, Integers)
self.assertIs(instance_cloned.x.domain, Integers)
def test_relax_integrality(self):
# Coverage of the _clear_attribute method
self.model.d = Var(within=Integers, bounds=(-2,3))
instance=self.model.create_instance()
instance_cloned = instance.clone()
xfrm = TransformationFactory('core.relax_integer_vars')
rinst = xfrm.create_using(instance_cloned)
self.assertEqual(type(rinst.d.domain), RealSet)
self.assertEqual(rinst.d.bounds, (-2,3))
self.assertIs(instance.d.domain, Integers)
self.assertIs(instance_cloned.d.domain, Integers)
def test_fix_discrete(self):
# Coverage of the _clear_attribute method
self.model.A = RangeSet(1,4)
self.model.a = Var()
self.model.b = Var(within=self.model.A)
self.model.c = Var(within=NonNegativeIntegers)
self.model.d = Var(within=Integers, bounds=(-2,3))
self.model.e = Var(within=Boolean)
self.model.f = Var(domain=Boolean)
instance=self.model.create_instance()
xfrm = TransformationFactory('core.fix_discrete')
rinst = xfrm.create_using(instance)
self.assertFalse(rinst.a.is_fixed())
self.assertTrue(rinst.b.is_fixed())
self.assertTrue(rinst.c.is_fixed())
self.assertTrue(rinst.d.is_fixed())
self.assertTrue(rinst.e.is_fixed())
self.assertTrue(rinst.f.is_fixed())
def test_nonnegative_transform_3(self):
self.model.S = RangeSet(0,10)
self.model.T = Set(initialize=["foo", "bar"])
# Unindexed, singly indexed, and doubly indexed variables with
# explicit bounds
self.model.x1 = Var(bounds=(-3, 3))
self.model.y1 = Var(self.model.S, bounds=(-3, 3))
self.model.z1 = Var(self.model.S, self.model.T, bounds=(-3, 3))
# Unindexed, singly indexed, and doubly indexed variables with
# rule-defined bounds
def boundsRule(*args):
return (-4, 4)
self.model.x2 = Var(bounds=boundsRule)
self.model.y2 = Var(self.model.S, bounds=boundsRule)
self.model.z2 = Var(self.model.S, self.model.T, bounds=boundsRule)
# Unindexed, singly indexed, and doubly indexed variables with
# explicit domains
self.model.x3 = Var(domain=NegativeReals, bounds=(-10, 10))
self.model.y3 = Var(self.model.S, domain = NegativeIntegers, bounds=(-10, 10))
self.model.z3 = Var(self.model.S, self.model.T, domain = Reals, bounds=(-10, 10))
# Unindexed, singly indexed, and doubly indexed variables with
# rule-defined domains
def domainRule(*args):
if len(args) == 1:
arg = 0
else:
arg = args[1]
if len(args) == 1 or arg == 0:
return NonNegativeReals
elif arg == 1:
return NonNegativeIntegers
elif arg == 2:
return NonPositiveReals
elif arg == 3:
return NonPositiveIntegers
elif arg == 4:
return NegativeReals
elif arg == 5:
return NegativeIntegers
elif arg == 6:
return PositiveReals
elif arg == 7:
return PositiveIntegers
elif arg == 8:
return Reals
elif arg == 9:
return Integers
elif arg == 10:
return Binary
else:
return Reals
self.model.x4 = Var(domain=domainRule, bounds=(-10, 10))
self.model.y4 = Var(self.model.S, domain=domainRule, bounds=(-10, 10))
self.model.z4 = Var(self.model.S, self.model.T, domain=domainRule, bounds=(-10, 10))
def objRule(model):
return sum(5*sum_product(model.__getattribute__(c+n)) \
for c in ('x', 'y', 'z') for n in ('1', '2', '3', '4'))
self.model.obj = Objective(rule=objRule)
transform = TransformationFactory('core.nonnegative_vars')
instance=self.model.create_instance()
transformed = transform.create_using(instance)
opt = SolverFactory("glpk")
instance_sol = opt.solve(instance)
transformed_sol = opt.solve(transformed)
self.assertEqual(
instance_sol["Solution"][0]["Objective"]['obj']["value"],
transformed_sol["Solution"][0]["Objective"]['obj']["value"]
)
def test_nonnegativity_transformation_2(self):
self.model.S = RangeSet(0,10)
self.model.T = Set(initialize=["foo", "bar"])
# Unindexed, singly indexed, and doubly indexed variables with
# explicit bounds
self.model.x1 = Var(bounds=(-3, 3))
self.model.y1 = Var(self.model.S, bounds=(-3, 3))
self.model.z1 = Var(self.model.S, self.model.T, bounds=(-3, 3))
# Unindexed, singly indexed, and doubly indexed variables with
# rule-defined bounds
def boundsRule(*args):
return (-4, 4)
self.model.x2 = Var(bounds=boundsRule)
self.model.y2 = Var(self.model.S, bounds=boundsRule)
self.model.z2 = Var(self.model.S, self.model.T, bounds=boundsRule)
# Unindexed, singly indexed, and doubly indexed variables with
# explicit domains
self.model.x3 = Var(domain=NegativeReals)
self.model.y3 = Var(self.model.S, domain = NegativeIntegers)
self.model.z3 = Var(self.model.S, self.model.T, domain = Reals)
# Unindexed, singly indexed, and doubly indexed variables with
# rule-defined domains
def domainRule(*args):
if len(args) == 1 or args[0] == 0:
return NonNegativeReals
elif args[0] == 1:
return NonNegativeIntegers
elif args[0] == 2:
return NonPositiveReals
elif args[0] == 3:
return NonPositiveIntegers
elif args[0] == 4:
return NegativeReals
elif args[0] == 5:
return NegativeIntegers
elif args[0] == 6:
return PositiveReals
elif args[0] == 7:
return PositiveIntegers
elif args[0] == 8:
return Reals
elif args[0] == 9:
return Integers
elif args[0] == 10:
return Binary
else:
return NonNegativeReals
self.model.x4 = Var(domain=domainRule)
self.model.y4 = Var(self.model.S, domain=domainRule)
self.model.z4 = Var(self.model.S, self.model.T, domain=domainRule)
instance = self.model.create_instance()
xfrm = TransformationFactory('core.nonnegative_vars')
transformed = xfrm.create_using(instance)
# Make sure everything is nonnegative
for c in ('x', 'y', 'z'):
for n in ('1', '2', '3', '4'):
var = transformed.__getattribute__(c+n)
for ndx in var._index:
self.assertTrue(self.nonnegativeBounds(var[ndx]))
