test_cvxpy.py•5.14 kB
from returns.result import Success
from usolver_mcp.models.cvxpy_models import (
CVXPYConstraint,
CVXPYObjective,
CVXPYProblem,
CVXPYVariable,
ObjectiveType,
)
from usolver_mcp.solvers.cvxpy_solver import solve_cvxpy_problem
pytest_plugins: list[str] = []
def test_simple_linear_regression() -> None:
"""Test solving a simple linear regression problem using CVXPY.
Problem: minimize ||Ax - b||²
where A = [[1, 1], [2, 1], [3, 1]] and b = [3, 5, 7]
Expected solution: x ≈ [2, 1] (slope=2, intercept=1)
"""
problem = CVXPYProblem(
variables=[CVXPYVariable(name="x", shape=2)],
objective=CVXPYObjective(
type=ObjectiveType.MINIMIZE,
expression="cp.sum_squares(np.array(A) @ x - np.array(b))",
),
constraints=[],
parameters={"A": [[1, 1], [2, 1], [3, 1]], "b": [3, 5, 7]},
description="Simple linear regression problem",
)
result = solve_cvxpy_problem(problem)
assert isinstance(result, Success)
solution = result.unwrap()
# Check that the problem was solved successfully
assert solution.status == "optimal"
assert solution.objective_value is not None
assert solution.objective_value < 1e-10 # Should be very close to 0
# Check solution values (should be approximately [2, 1])
x_values = solution.values["x"]
assert abs(x_values[0] - 2.0) < 1e-6 # slope
assert abs(x_values[1] - 1.0) < 1e-6 # intercept
def test_constrained_optimization() -> None:
"""Test CVXPY with constraints.
Problem: minimize x² + y²
subject to: x + y >= 2, x >= 0, y >= 0
Expected solution: x = 1, y = 1, objective = 2
"""
problem = CVXPYProblem(
variables=[CVXPYVariable(name="x", shape=1), CVXPYVariable(name="y", shape=1)],
objective=CVXPYObjective(
type=ObjectiveType.MINIMIZE,
expression="cp.sum_squares(x) + cp.sum_squares(y)",
),
constraints=[
CVXPYConstraint(expression="x + y >= 2"),
CVXPYConstraint(expression="x >= 0"),
CVXPYConstraint(expression="y >= 0"),
],
description="Constrained quadratic optimization",
)
result = solve_cvxpy_problem(problem)
assert isinstance(result, Success)
solution = result.unwrap()
assert solution.status == "optimal"
assert solution.objective_value is not None
# Check solution values (should be approximately x=1, y=1)
x_val = (
solution.values["x"][0]
if hasattr(solution.values["x"], "__len__")
else solution.values["x"]
)
y_val = (
solution.values["y"][0]
if hasattr(solution.values["y"], "__len__")
else solution.values["y"]
)
assert abs(x_val - 1.0) < 1e-4
assert abs(y_val - 1.0) < 1e-4
assert abs(solution.objective_value - 2.0) < 1e-4
def test_portfolio_optimization() -> None:
"""Test a simple portfolio optimization problem.
Problem: minimize risk while achieving target return
Two assets with different risk/return profiles
"""
problem = CVXPYProblem(
variables=[CVXPYVariable(name="weights", shape=2)],
objective=CVXPYObjective(
type=ObjectiveType.MINIMIZE, expression="cp.quad_form(weights, cov_matrix)"
),
constraints=[
CVXPYConstraint(expression="cp.sum(weights) == 1"), # weights sum to 1
CVXPYConstraint(expression="weights >= 0"), # long-only
CVXPYConstraint(
expression="np.array(returns).T @ weights >= target_return"
), # target return
],
parameters={
"cov_matrix": [[0.04, 0.01], [0.01, 0.09]], # covariance matrix
"returns": [0.08, 0.12], # expected returns
"target_return": 0.10, # target 10% return
},
description="Simple portfolio optimization",
)
result = solve_cvxpy_problem(problem)
assert isinstance(result, Success)
solution = result.unwrap()
assert solution.status == "optimal"
assert solution.objective_value is not None
# Check that weights sum to 1
weights = solution.values["weights"]
assert abs(sum(weights) - 1.0) < 1e-6
# Check that all weights are non-negative
assert all(w >= -1e-6 for w in weights)
def test_infeasible_problem() -> None:
"""Test CVXPY with an infeasible problem.
Problem: minimize x
subject to: x >= 2, x <= 1
This should be infeasible.
"""
problem = CVXPYProblem(
variables=[CVXPYVariable(name="x", shape=1)],
objective=CVXPYObjective(type=ObjectiveType.MINIMIZE, expression="x"),
constraints=[
CVXPYConstraint(expression="x >= 2"),
CVXPYConstraint(expression="x <= 1"),
],
description="Infeasible optimization problem",
)
result = solve_cvxpy_problem(problem)
assert isinstance(result, Success)
solution = result.unwrap()
# Should be infeasible
assert solution.status == "infeasible"
assert solution.objective_value is None or solution.objective_value == float("inf")