def test_addition_of_var_and_constant(expr):
m = vlin.Model(expr=expr)
a = m.var(2)
v = a + 2.34
assert v.shape == a.shape
assert np.allclose(v.rawdense()[:, 0], 2.34)
assert isinstance(v, expr)
v = a + np.array([2.34])
assert v.shape == a.shape
assert np.allclose(v.rawdense()[:, 0], 2.34)
assert isinstance(v, expr)
v = a + np.array([2.34, 1.23])
assert v.shape == a.shape
assert np.allclose(v.rawdense()[:, 0], [2.34, 1.23])
assert isinstance(v, expr)
with pytest.raises(ValueError) as execinfo:
a + np.array([1, 2, 3])
assert "operands could not be broadcast together" in str(execinfo)
with pytest.raises(NotImplementedError):
a += 1
def test_multiplication_of_var_and_constant(expr):
m = vlin.Model(expr=expr)
a = m.var(2)
v = a * 2.34
assert v.shape == a.shape
assert np.allclose(v.rawdense().sum(axis=1), 2.34)
assert isinstance(v, expr)
v = a * np.array([2.34])
assert v.shape == a.shape
assert np.allclose(v.rawdense().sum(axis=1), 2.34)
assert isinstance(v, expr)
v = a * np.array([2.34, 1.23])
assert v.shape == a.shape
assert np.allclose(v.rawdense().sum(axis=1), [2.34, 1.23])
assert isinstance(v, expr)
with pytest.raises(ValueError) as execinfo:
a * np.array([1, 2, 3])
assert "operands could not be broadcast together" in str(
execinfo) or "inconsistent shapes" in str(execinfo)
with pytest.raises(NotImplementedError):
a *= 1
def test_trivial_fully_linear(expr, solver):
m = vlin.Model(expr=expr, max_vars=4)
a = m.var(3)
m += a <= np.array([1.1, 2.3, 3.3])
m += a >= 0
m.objective = -1 * a.sum() # Maximise sum of them.
res, x = m.solve(solver=solver)
assert np.allclose(a.raw() @ x, np.array([1.1, 2.3, 3.3]))
def test_trivial_milp(expr, solver):
m = vlin.Model(expr=expr, max_vars=4)
a = m.var(3, integer=np.array([False, True, False]))
m += a <= np.array([1.1, 2.3, 3.3])
m += a >= 0
m.objective = -1 * a.sum() # Maximise sum of them.
res, x = m.solve(solver=solver)
assert np.allclose(a.raw() @ x, np.array([1.1, 2.0, 3.3]))
def test_sum(expr):
m = vlin.Model(expr=expr, max_vars=4)
a = m.var(1)
b = m.var(3)
assert np.allclose(a.sum().rawdense(), [[0, 1, 0, 0, 0]])
# Broadcasting~~
# This is like [0, 1, 2] + 1 => [1, 2, 3]
# and then summed: [1, 2, 3].sum() = 6 == [0, 1, 2].sum() * 1*len([0, 1, 2])
assert np.allclose((a + b).sum().rawdense(), [[0, 3, 1, 1, 1]])
assert a.sum().shape == (1, m.max_vars)
assert b.sum().shape == (1, m.max_vars)
def test_knapsack(expr, solver):
# https://docs.python-mip.com/en/latest/examples.html
m = vlin.Model(expr=expr, max_vars=20)
p = np.array([10, 13, 18, 31, 7, 15])
w = np.array([11, 15, 20, 35, 10, 33])
c = 47
x = m.var(len(w), integer=True)
m += x >= 0
m += x <= 1
m.objective = -1 * (x * p).sum()
m += (x * w).sum() <= c
res, sol_x = m.solve(solver=solver)
print(x.raw() @ sol_x)
assert np.isclose(-41, res["value"])
def test_addition_of_var_and_var(expr):
m = vlin.Model(expr=expr, max_vars=8)
a = m.var(2)
b = m.var(2)
c = m.var(3)
d = m.var(1)
assert np.allclose((a + b).rawdense(), a.rawdense() + b.rawdense())
# Force broadcasting.
assert np.allclose((a + d).rawdense(), a.rawdense() + d.rawdense())
assert np.allclose((d + a).rawdense(), a.rawdense() + d.rawdense())
with pytest.raises(ValueError) as execinfo:
a + c
assert "operands could not be broadcast together" in str(
execinfo) or "inconsistent shapes" in str(execinfo)
def test_slicing_shape(expr):
m = vlin.Model(expr=expr)
a = m.var(4)
assert a[1:3].shape == (2, m.max_vars)
assert a[1].shape == (1, m.max_vars)
def test_slicing_type(expr):
m = vlin.Model(expr=expr)
a = m.var(4)
assert isinstance(a[1:3], expr)
assert isinstance(a[1], expr)