""" Copyright 2019, the CVXPY developers. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. """ import math import re import unittest import numpy as np import pytest import scipy.linalg as la import scipy.stats as st import cvxpy as cp import cvxpy.tests.solver_test_helpers as sths from cvxpy.reductions.solvers.defines import ( INSTALLED_MI_SOLVERS, INSTALLED_SOLVERS, ) from cvxpy.tests.base_test import BaseTest from cvxpy.tests.solver_test_helpers import ( StandardTestECPs, StandardTestLPs, StandardTestMixedCPs, StandardTestPCPs, StandardTestQPs, StandardTestSDPs, StandardTestSOCPs, ) from cvxpy.utilities.versioning import Version @unittest.skipUnless('ECOS' in INSTALLED_SOLVERS, 'ECOS is not installed.') class TestECOS(BaseTest): def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_ecos_options(self) -> None: """Test that all the ECOS solver options work. """ # Test ecos # feastol, abstol, reltol, feastol_inacc, # abstol_inacc, and reltol_inacc for tolerance values # max_iters for the maximum number of iterations, EPS = 1e-4 prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1) + 1.0), [self.x == 0]) for i in range(2): prob.solve(solver=cp.ECOS, feastol=EPS, abstol=EPS, reltol=EPS, feastol_inacc=EPS, abstol_inacc=EPS, reltol_inacc=EPS, max_iters=20, verbose=True, warm_start=True) self.assertAlmostEqual(prob.value, 1.0) self.assertItemsAlmostEqual(self.x.value, [0, 0]) def test_ecos_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='ECOS') def test_ecos_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='ECOS') def test_ecos_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='ECOS') def test_ecos_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='ECOS') def test_ecos_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='ECOS') def test_ecos_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='ECOS') def test_ecos_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='ECOS') def test_ecos_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='ECOS') def test_ecos_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='ECOS') def test_ecos_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='ECOS') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='ECOS') def test_ecos_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='ECOS') def test_ecos_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='ECOS') class TestSCS(BaseTest): """ Unit tests for SCS. """ def setUp(self) -> None: self.x = cp.Variable(2, name='x') self.y = cp.Variable(2, name='y') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') # Overridden method to assume lower accuracy. def assertItemsAlmostEqual(self, a, b, places: int = 2) -> None: super(TestSCS, self).assertItemsAlmostEqual(a, b, places=places) # Overridden method to assume lower accuracy. def assertAlmostEqual(self, a, b, places: int = 2) -> None: super(TestSCS, self).assertAlmostEqual(a, b, places=places) def test_scs_retry(self) -> None: """Test that SCS retry doesn't trigger a crash. """ n_sec = 20 np.random.seed(315) mu = np.random.random(n_sec) random_mat = np.random.rand(n_sec, n_sec) C = np.dot(random_mat, random_mat.transpose()) x = cp.Variable(n_sec) prob = cp.Problem(cp.Minimize(cp.QuadForm(x, C)), [cp.sum(x) == 1, 0 <= x, x <= 1, x @ mu >= np.max(mu) - 1e-6]) prob.solve(cp.SCS) assert prob.status in {cp.OPTIMAL, cp.OPTIMAL_INACCURATE} def test_scs_options(self) -> None: """Test that all the SCS solver options work. """ # Test SCS # MAX_ITERS, EPS, ALPHA, UNDET_TOL, VERBOSE, and NORMALIZE. # If opts is missing, then the algorithm uses default settings. # USE_INDIRECT = True EPS = 1e-4 x = cp.Variable(2, name='x') prob = cp.Problem(cp.Minimize(cp.norm(x, 1) + 1.0), [x == 0]) for i in range(2): prob.solve(solver=cp.SCS, max_iters=50, eps=EPS, alpha=1.2, verbose=True, normalize=True, use_indirect=False) self.assertAlmostEqual(prob.value, 1.0, places=2) self.assertItemsAlmostEqual(x.value, [0, 0], places=2) def test_log_problem(self) -> None: # Log in objective. obj = cp.Maximize(cp.sum(cp.log(self.x))) constr = [self.x <= [1, math.e]] p = cp.Problem(obj, constr) result = p.solve(solver=cp.SCS) self.assertAlmostEqual(result, 1) self.assertItemsAlmostEqual(self.x.value, [1, math.e]) # Log in constraint. obj = cp.Minimize(sum(self.x)) constr = [cp.log(self.x) >= 0, self.x <= [1, 1]] p = cp.Problem(obj, constr) result = p.solve(solver=cp.SCS) self.assertAlmostEqual(result, 2) self.assertItemsAlmostEqual(self.x.value, [1, 1]) # Index into log. obj = cp.Maximize(cp.log(self.x)[1]) constr = [self.x <= [1, math.e]] p = cp.Problem(obj, constr) result = p.solve(solver=cp.SCS) def test_sigma_max(self) -> None: """Test sigma_max. """ const = cp.Constant([[1, 2, 3], [4, 5, 6]]) constr = [self.C == const] prob = cp.Problem(cp.Minimize(cp.norm(self.C, 2)), constr) result = prob.solve(solver=cp.SCS) self.assertAlmostEqual(result, cp.norm(const, 2).value) self.assertItemsAlmostEqual(self.C.value, const.value) def test_sdp_var(self) -> None: """Test sdp var. """ const = cp.Constant([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) X = cp.Variable((3, 3), PSD=True) prob = cp.Problem(cp.Minimize(0), [X == const]) prob.solve(solver=cp.SCS) self.assertEqual(prob.status, cp.INFEASIBLE) def test_complex_matrices(self) -> None: """Test complex matrices. """ # Complex-valued matrix np.random.seed(0) K = np.array(np.random.rand(2, 2) + 1j * np.random.rand(2, 2)) # example matrix n1 = la.svdvals(K).sum() # trace norm of K # Dual Problem X = cp.Variable((2, 2), complex=True) Y = cp.Variable((2, 2), complex=True) # X, Y >= 0 so trace is real objective = cp.Minimize( cp.real(0.5 * cp.trace(X) + 0.5 * cp.trace(Y)) ) constraints = [ cp.bmat([[X, -K.conj().T], [-K, Y]]) >> 0, X >> 0, Y >> 0, ] problem = cp.Problem(objective, constraints) sol_scs = problem.solve(solver='SCS') self.assertEqual(constraints[0].dual_value.shape, (4, 4)) self.assertEqual(constraints[1].dual_value.shape, (2, 2)) self.assertEqual(constraints[2].dual_value.shape, (2, 2)) self.assertAlmostEqual(sol_scs, n1) def test_entr(self) -> None: """Test a problem with entr. """ for n in [5, 10, 25]: print(n) x = cp.Variable(n) obj = cp.Maximize(cp.sum(cp.entr(x))) p = cp.Problem(obj, [cp.sum(x) == 1]) p.solve(solver=cp.SCS) self.assertItemsAlmostEqual(x.value, n*[1./n]) def test_exp(self) -> None: """Test a problem with exp. """ for n in [5, 10, 25]: print(n) x = cp.Variable(n) obj = cp.Minimize(cp.sum(cp.exp(x))) p = cp.Problem(obj, [cp.sum(x) == 1]) p.solve(solver=cp.SCS) self.assertItemsAlmostEqual(x.value, n*[1./n]) def test_log(self) -> None: """Test a problem with log. """ for n in [5, 10, 25]: print(n) x = cp.Variable(n) obj = cp.Maximize(cp.sum(cp.log(x))) p = cp.Problem(obj, [cp.sum(x) == 1]) p.solve(solver=cp.SCS) self.assertItemsAlmostEqual(x.value, n*[1./n]) def test_solve_problem_twice(self) -> None: """Test a problem with log. """ n = 5 x = cp.Variable(n) obj = cp.Maximize(cp.sum(cp.log(x))) p = cp.Problem(obj, [cp.sum(x) == 1]) p.solve(solver=cp.SCS) first_value = x.value self.assertItemsAlmostEqual(first_value, n*[1./n]) p.solve(solver=cp.SCS) second_value = x.value self.assertItemsAlmostEqual(first_value, second_value) def test_warm_start(self) -> None: """Test warm starting. """ x = cp.Variable(10) obj = cp.Minimize(cp.sum(cp.exp(x))) prob = cp.Problem(obj, [cp.sum(x) == 1]) result = prob.solve(solver=cp.SCS) time = prob.solver_stats.solve_time result2 = prob.solve(solver=cp.SCS, warm_start=True) time2 = prob.solver_stats.solve_time self.assertAlmostEqual(result2, result, places=2) print(time > time2) def test_warm_start_diffcp(self) -> None: """Test warm starting in diffcvx. """ try: import diffcp diffcp # for flake8 except ModuleNotFoundError: self.skipTest("diffcp not installed.") x = cp.Variable(10) obj = cp.Minimize(cp.sum(cp.exp(x))) prob = cp.Problem(obj, [cp.sum(x) == 1]) result = prob.solve(solver=cp.DIFFCP) result2 = prob.solve(solver=cp.DIFFCP, warm_start=True) self.assertAlmostEqual(result2, result, places=2) def test_psd_constraint(self) -> None: """Test PSD constraint. """ s = cp.Variable((2, 2)) obj = cp.Maximize(cp.minimum(s[0, 1], 10)) const = [s >> 0, cp.diag(s) == np.ones(2)] prob = cp.Problem(obj, const) r = prob.solve(solver=cp.SCS) s = s.value print(const[0].residual) print("value", r) print("s", s) print("eigs", np.linalg.eig(s + s.T)[0]) eigs = np.linalg.eig(s + s.T)[0] self.assertEqual(np.all(eigs >= 0), True) def test_quad_obj(self) -> None: """Test SCS canonicalization with a quadratic objective. """ # Only relevant for SCS >= 3.0.0. import scs if Version(scs.__version__) >= Version('3.0.0'): x = cp.Variable(2) expr = cp.sum_squares(x) constr = [x >= 1] prob = cp.Problem(cp.Minimize(expr), constr) data = prob.get_problem_data(solver=cp.SCS) self.assertItemsAlmostEqual(data[0]["P"].toarray(), 2*np.eye(2)) solution1 = prob.solve(solver=cp.SCS) # When use_quad_obj = False, the quadratic objective is # canonicalized to a SOC constraint. prob = cp.Problem(cp.Minimize(expr), constr) solver_opts = {"use_quad_obj": False} data = prob.get_problem_data(solver=cp.SCS, solver_opts=solver_opts) assert "P" not in data[0] solution2 = prob.solve(solver=cp.SCS, **solver_opts) assert np.isclose(solution1, solution2) # Check that there is no P for non-quadratic objectives. expr = cp.norm(x, 1) prob = cp.Problem(cp.Minimize(expr), constr) data = prob.get_problem_data(solver=cp.SCS) assert "P" not in data[0] def test_quad_obj_with_power(self) -> None: """Test a mixed quadratic/power objective. """ # Only relevant for SCS >= 3.0.0. import scs if Version(scs.__version__) >= Version('3.0.0'): # See https://github.com/cvxpy/cvxpy/issues/2059 x = cp.Variable() prob = cp.Problem(cp.Minimize(x**1.6 + x**2), [x >= 1]) prob.solve(solver=cp.SCS, use_quad_obj=True) self.assertAlmostEqual(prob.value, 2) self.assertAlmostEqual(x.value, 1) # Check problem data. data = prob.get_problem_data(solver=cp.SCS, solver_opts={"use_quad_obj": True}) # Quadratic objective and SOC contraints. assert "P" in data[0] assert data[0]["dims"].soc def test_scs_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='SCS') def test_scs_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='SCS') def test_scs_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='SCS', eps=1e-6) def test_scs_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='SCS', eps=1e-6) def test_scs_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='SCS') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='SCS') def test_scs_sdp_1min(self) -> None: StandardTestSDPs.test_sdp_1min(solver='SCS') def test_scs_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver='SCS', eps=1e-5) def test_scs_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='SCS', eps=1e-5) def test_scs_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='SCS', eps=1e-5) def test_scs_sdp_pcp_1(self): StandardTestMixedCPs.test_sdp_pcp_1(solver='SCS') def test_scs_pcp_1(self) -> None: StandardTestPCPs.test_pcp_1(solver='SCS') def test_scs_pcp_2(self) -> None: StandardTestPCPs.test_pcp_2(solver='SCS') def test_scs_pcp_3(self) -> None: StandardTestPCPs.test_pcp_3(solver='SCS', eps=1e-12) @unittest.skipUnless('CLARABEL' in INSTALLED_SOLVERS, 'CLARABEL is not installed.') class TestClarabel(BaseTest): """ Unit tests for Clarabel. """ def setUp(self) -> None: self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_clarabel_parameter_update(self) -> None: """Test warm start. """ x = cp.Variable(2) P = cp.Parameter(nonneg=True), A = cp.Parameter(4) b = cp.Parameter(2, nonneg=True) q = cp.Parameter(2) def update_parameters(P, A, b, q): P[0].value = np.random.rand() A.value = np.random.randn(4) b.value = np.random.rand(2) q.value = np.random.randn(2) prob = cp.Problem( cp.Minimize(P[0]*cp.square(x[0]) + cp.quad_form(x, np.ones([2, 2])) + q.T @ x), [A[0] * x[0] + A[1] * x[1] == b[0], A[2] * x[0] + A[3] * x[1] <= b[1]] ) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.CLARABEL, warm_start=False) result2 = prob.solve(solver=cp.CLARABEL, warm_start=True) self.assertAlmostEqual(result1, result2) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.CLARABEL, warm_start=True) result2 = prob.solve(solver=cp.CLARABEL, warm_start=False) self.assertAlmostEqual(result1, result2) # consecutive solves, no data update result1 = prob.solve(solver=cp.CLARABEL, warm_start=False) self.assertAlmostEqual(result1, result2) def test_clarabel_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver=cp.CLARABEL) def test_clarabel_nonstandard_name(self) -> None: # Test that solver name with non-standard capitalization works. StandardTestLPs.test_lp_0(solver="Clarabel") def test_clarabel_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='CLARABEL') def test_clarabel_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='CLARABEL') def test_clarabel_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='CLARABEL') def test_clarabel_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='CLARABEL') def test_clarabel_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='CLARABEL') def test_clarabel_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver='CLARABEL') def test_clarabel_qp_0_linear_obj(self) -> None: StandardTestQPs.test_qp_0(solver='CLARABEL', use_quad_obj=False) def test_clarabel_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CLARABEL') def test_clarabel_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='CLARABEL') def test_clarabel_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CLARABEL') def test_clarabel_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='CLARABEL') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='CLARABEL') def test_clarabel_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='CLARABEL') def test_clarabel_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='CLARABEL') def test_clarabel_pcp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CLARABEL') def test_clarabel_pcp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='CLARABEL') def test_clarabel_pcp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CLARABEL') def test_clarabel_sdp_1min(self) -> None: StandardTestSDPs.test_sdp_1min(solver='CLARABEL') def test_clarabel_sdp_2(self) -> None: # produces a different optimizer than # the one expected by the standard test places = 3 sth = sths.sdp_2() sth.solve('CLARABEL') sth.verify_objective(places) sth.check_primal_feasibility(places) # sth.verify_primal_values(places) # skip sth.check_complementarity(places) sth.check_dual_domains(places) @unittest.skipUnless('CUCLARABEL' in INSTALLED_SOLVERS, 'CLARABEL is not installed.') class TestCuClarabel(BaseTest): """ Unit tests for Clarabel. """ def setUp(self) -> None: self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_clarabel_parameter_update(self) -> None: """Test warm start. """ x = cp.Variable(2) P = cp.Parameter(nonneg=True), A = cp.Parameter(4) b = cp.Parameter(2, nonneg=True) q = cp.Parameter(2) def update_parameters(P, A, b, q): P[0].value = np.random.rand() A.value = np.random.randn(4) b.value = np.random.rand(2) q.value = np.random.randn(2) prob = cp.Problem( cp.Minimize(P[0]*cp.square(x[0]) + cp.quad_form(x, np.ones([2, 2])) + q.T @ x), [A[0] * x[0] + A[1] * x[1] == b[0], A[2] * x[0] + A[3] * x[1] <= b[1]] ) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.CLARABEL, warm_start=False) result2 = prob.solve(solver=cp.CLARABEL, warm_start=True) self.assertAlmostEqual(result1, result2) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.CLARABEL, warm_start=True) result2 = prob.solve(solver=cp.CLARABEL, warm_start=False) self.assertAlmostEqual(result1, result2) # consecutive solves, no data update result1 = prob.solve(solver=cp.CLARABEL, warm_start=False) self.assertAlmostEqual(result1, result2) def test_clarabel_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver=cp.CUCLARABEL) def test_clarabel_nonstandard_name(self) -> None: # Test that solver name with non-standard capitalization works. StandardTestLPs.test_lp_0(solver="CuClarabel") def test_clarabel_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='CUCLARABEL') def test_clarabel_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='CUCLARABEL') def test_clarabel_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='CUCLARABEL') def test_clarabel_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='CUCLARABEL') def test_clarabel_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='CUCLARABEL') def test_clarabel_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver='CUCLARABEL') def test_clarabel_qp_0_linear_obj(self) -> None: StandardTestQPs.test_qp_0(solver='CUCLARABEL', use_quad_obj=False) def test_clarabel_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CUCLARABEL') def test_clarabel_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='CUCLARABEL') def test_clarabel_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CUCLARABEL') def test_clarabel_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='CUCLARABEL') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='CUCLARABEL') def test_clarabel_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='CUCLARABEL') def test_clarabel_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='CUCLARABEL') def test_clarabel_pcp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CUCLARABEL') def test_clarabel_pcp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='CUCLARABEL') def test_clarabel_pcp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CUCLARABEL') @unittest.skipUnless('MOREAU' in INSTALLED_SOLVERS, 'MOREAU is not installed.') class TestMoreau(BaseTest): """ Unit tests for Moreau. """ def setUp(self) -> None: self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_moreau_parameter_update(self) -> None: """Test warm start. """ x = cp.Variable(2) P = cp.Parameter(nonneg=True), A = cp.Parameter(4) b = cp.Parameter(2, nonneg=True) q = cp.Parameter(2) def update_parameters(P, A, b, q): P[0].value = np.random.rand() A.value = np.random.randn(4) b.value = np.random.rand(2) q.value = np.random.randn(2) prob = cp.Problem( cp.Minimize(P[0]*cp.square(x[0]) + cp.quad_form(x, np.ones([2, 2])) + q.T @ x), [A[0] * x[0] + A[1] * x[1] == b[0], A[2] * x[0] + A[3] * x[1] <= b[1]] ) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.MOREAU, warm_start=False) result2 = prob.solve(solver=cp.MOREAU, warm_start=True) self.assertAlmostEqual(result1, result2) update_parameters(P, A, b, q) result1 = prob.solve(solver=cp.MOREAU, warm_start=True) result2 = prob.solve(solver=cp.MOREAU, warm_start=False) self.assertAlmostEqual(result1, result2) # consecutive solves, no data update result1 = prob.solve(solver=cp.MOREAU, warm_start=False) self.assertAlmostEqual(result1, result2) def test_moreau_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver=cp.MOREAU) def test_moreau_nonstandard_name(self) -> None: # Test that solver name with non-standard capitalization works. StandardTestLPs.test_lp_0(solver="MOREAU") def test_moreau_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='MOREAU') def test_moreau_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='MOREAU') def test_moreau_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='MOREAU') def test_moreau_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='MOREAU') def test_moreau_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='MOREAU') def test_moreau_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver='MOREAU') def test_moreau_qp_0_linear_obj(self) -> None: StandardTestQPs.test_qp_0(solver='MOREAU', use_quad_obj=False) def test_moreau_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='MOREAU') def test_moreau_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='MOREAU') def test_moreau_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='MOREAU') def test_moreau_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='MOREAU') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='MOREAU') def test_moreau_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='MOREAU') def test_moreau_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='MOREAU') def test_moreau_pcp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='MOREAU') def test_moreau_pcp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='MOREAU') def test_moreau_pcp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='MOREAU') def is_mosek_available(): """Check if MOSEK is installed and a license is available.""" if 'MOSEK' not in INSTALLED_SOLVERS: return False try: import mosek # type: ignore env = mosek.Env() # Try to get license status (returns 0 if OK) status = env.getlicense() return status == mosek.rescode.ok except Exception: return False @unittest.skipUnless(is_mosek_available(), 'MOSEK is not installed or license is not available.') class TestMosek(unittest.TestCase): def test_mosek_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='MOSEK') def test_mosek_lp_1(self) -> None: # default settings StandardTestLPs.test_lp_1(solver='MOSEK') # require a basic feasible solution StandardTestLPs.test_lp_1(solver='MOSEK', places=6, bfs=True) def test_mosek_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='MOSEK') def test_mosek_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='MOSEK') def test_mosek_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='MOSEK') def test_mosek_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='MOSEK') def test_mosek_lp_bound_attr(self) -> None: StandardTestLPs.test_lp_bound_attr(solver='MOSEK') def test_mosek_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='MOSEK') def test_mosek_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='MOSEK') def test_mosek_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='MOSEK') def test_mosek_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='MOSEK') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='MOSEK') def test_mosek_sdp_1(self) -> None: # minimization StandardTestSDPs.test_sdp_1min(solver='MOSEK') # maximization StandardTestSDPs.test_sdp_1max(solver='MOSEK') def test_mosek_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver='MOSEK') def test_mosek_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='MOSEK') def test_mosek_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='MOSEK') def test_mosek_pcp_1(self) -> None: StandardTestPCPs.test_pcp_1(solver='MOSEK', places=2) def test_mosek_pcp_2(self) -> None: StandardTestPCPs.test_pcp_2(solver='MOSEK') def test_mosek_pcp_3(self) -> None: StandardTestPCPs.test_pcp_3(solver='MOSEK') def test_mosek_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='MOSEK') def test_mosek_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='MOSEK') def test_mosek_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='MOSEK') def test_mosek_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='MOSEK') def test_mosek_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='MOSEK') def test_mosek_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='MOSEK', places=3) def test_mosek_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver='MOSEK') def test_mosek_mi_pcp_0(self) -> None: StandardTestPCPs.test_mi_pcp_0(solver='MOSEK') def test_mosek_params(self) -> None: import mosek n = 10 m = 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) invalid_mosek_params = { "MSK_IPAR_NUM_THREADS": "11.3" } with self.assertRaises(mosek.Error): problem.solve(solver=cp.MOSEK, mosek_params=invalid_mosek_params) with self.assertRaises(ValueError): problem.solve(solver=cp.MOSEK, invalid_kwarg=None) mosek_params = { mosek.dparam.basis_tol_x: 1e-8, "MSK_IPAR_INTPNT_MAX_ITERATIONS": 20, "MSK_IPAR_NUM_THREADS": "17", "MSK_IPAR_PRESOLVE_USE": "MSK_PRESOLVE_MODE_OFF", "MSK_DPAR_INTPNT_CO_TOL_DFEAS": 1e-9, "MSK_DPAR_INTPNT_CO_TOL_PFEAS": "1e-9" } with pytest.warns(): problem.solve(solver=cp.MOSEK, mosek_params=mosek_params) def test_mosek_simplex(self) -> None: n = 10 m = 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) problem.solve( solver=cp.MOSEK, mosek_params={"MSK_IPAR_OPTIMIZER": "MSK_OPTIMIZER_DUAL_SIMPLEX"} ) def test_mosek_iis(self) -> None: """Test IIS feature in Mosek.""" # Simple infeasible problem. n = 2 x = cp.Variable(n) objective = cp.Minimize(cp.sum(x)) constraints = [x[0] >= 1, x[0] <= -1, x[1] >= 3] problem = cp.Problem(objective, constraints) problem.solve(solver=cp.MOSEK) iis = problem.solver_stats.extra_stats["IIS"] assert iis[constraints[0].id] > 0 assert iis[constraints[1].id] > 0 assert iis[constraints[2].id] == 0 # More complicated infeasible problem. n = 3 m = 2 X = cp.Variable((m, n)) y = cp.Variable() objective = cp.Minimize(cp.sum(X)) constraints = [y == 2, X >= 3, X[0, 0] + y <= -5] problem = cp.Problem(objective, constraints) problem.solve(solver=cp.MOSEK) iis = problem.solver_stats.extra_stats["IIS"] assert abs(iis[constraints[0].id]) > 0 dual1 = np.reshape(iis[constraints[1].id], X.shape, order="C") assert dual1[0, 0] > 0 assert dual1[0, 1] == 0 assert np.all(dual1[1, :] == 0) assert iis[constraints[2].id] > 0 def test_mosek_sdp_power(self) -> None: """Test the problem in issue #2128""" StandardTestMixedCPs.test_sdp_pcp_1(solver='MOSEK') def test_power_portfolio(self) -> None: """Test the portfolio problem in issue #2042""" T, N = 200, 10 rs = np.random.RandomState(123) mean = np.zeros(N) + 1/1000 cov = rs.rand(N, N) * 1.5 - 0.5 cov = cov @ cov.T/1000 + np.diag(rs.rand(N) * 0.7 + 0.3)/1000 Y = st.multivariate_normal.rvs( mean=mean, cov=cov, size=T, random_state=rs ) w = cp.Variable((N, 1)) t = cp.Variable((1, 1)) z = cp.Variable((1, 1)) omega = cp.Variable((T, 1)) psi = cp.Variable((T, 1)) nu = cp.Variable((T, 1)) epsilon = cp.Variable((T, 1)) k = cp.Variable((1, 1)) b = np.ones((1, N))/N X = Y @ w h = 0.2 ones = np.ones((T, 1)) constraints = [ cp.constraints.power.PowCone3D(z * (1+h)/(2*h) * ones, psi * (1+h)/h, epsilon, 1/(1+h)), cp.constraints.power.PowCone3D(omega/(1-h), nu/h, -z/(2*h) * ones, (1-h)), -X - t + epsilon + omega <= 0, w >= 0, z >= 0, ] obj = t + z + cp.sum(psi + nu) constraints += [cp.sum(w) == k, k >= 0, b @ cp.log(w) >= 1, ] objective = cp.Minimize(obj) prob = cp.Problem(objective, constraints) prob.solve(solver=cp.MOSEK) assert prob.status is cp.OPTIMAL def test_mosek_accept_unknown(self) -> None: mosek_param = { "MSK_IPAR_INTPNT_MAX_ITERATIONS": 0 } sth = sths.lp_5() sth.solve(solver=cp.MOSEK, accept_unknown=True, mosek_params=mosek_param) assert sth.prob.status in {cp.OPTIMAL_INACCURATE, cp.OPTIMAL} with pytest.raises(cp.error.SolverError, match="Solver 'MOSEK' failed"): sth.solve(solver=cp.MOSEK, mosek_params=mosek_param) def test_mosek_number_iters(self) -> None: sth = sths.lp_5() sth.solve(solver=cp.MOSEK) assert sth.prob.solver_stats.num_iters >= 0 assert sth.prob.solver_stats.extra_stats["mio_intpnt_iter"] == 0 assert sth.prob.solver_stats.extra_stats["mio_simplex_iter"] == 0 def test_eps_keyword(self) -> None: """Test that the eps keyword is accepted""" x = cp.Variable() prob = cp.Problem(cp.Minimize(x), [x >= 0]) # This should not raise an exception prob.solve(solver=cp.MOSEK, eps=1e-8, mosek_params={'MSK_DPAR_INTPNT_CO_TOL_DFEAS': 1e-6}) assert prob.status is cp.OPTIMAL # This exception being raised shows that the eps value is being passed to MOSEK import mosek with pytest.raises(mosek.Error, match="The parameter value 0.1 is too large"): prob.solve(solver=cp.MOSEK, eps=1e-1, mosek_params={'MSK_DPAR_INTPNT_CO_TOL_DFEAS': 1e-6}) # If parameters are defined explicitly, eps will not overwrite -> no exception from cvxpy.reductions.solvers.conic_solvers.mosek_conif import MOSEK all_params = MOSEK.tolerance_params() prob.solve(solver=cp.MOSEK, eps=1e-1, mosek_params={p: 1e-6 for p in all_params}) assert prob.status is cp.OPTIMAL # Fails when used with enums with pytest.raises(AssertionError, match="not compatible"): prob.solve(solver=cp.MOSEK, eps=1e-1, mosek_params={mosek.dparam.intpnt_co_tol_dfeas: 1e-6}) @unittest.skipUnless('CVXOPT' in INSTALLED_SOLVERS, 'CVXOPT is not installed.') class TestCVXOPT(BaseTest): def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_cvxopt_options(self) -> None: """Test that all the CVXOPT solver options work. """ # 'maxiters' # maximum number of iterations (default: 100). # 'abstol' # absolute accuracy (default: 1e-7). # 'reltol' # relative accuracy (default: 1e-6). # 'feastol' # tolerance for feasibility conditions (default: 1e-7). # 'refinement' # number of iterative refinement steps when solving KKT equations # (default: 0 if the problem has no second-order cone # or matrix inequality constraints; 1 otherwise). EPS = 1e-7 prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1) + 1.0), [self.x == 0]) prob.solve(solver=cp.CVXOPT, feastol=EPS, abstol=EPS, reltol=EPS, max_iters=20, verbose=True, kktsolver='chol', refinement=2) self.assertAlmostEqual(prob.value, 1.0) self.assertItemsAlmostEqual(self.x.value, [0, 0]) msg = 'This setup-factor function was called.' def setup_dummy_factor(c, G, h, dims, A, b): # see cvxpy/reductions/solvers/kktsolver.py for an actual implementation # of a setup-factor function. raise NotImplementedError(msg) with self.assertRaises(NotImplementedError) as nix: prob.solve(solver='CVXOPT', kktsolver=setup_dummy_factor) self.assertEqual(msg, str(nix.exception)) def test_cvxopt_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='CVXOPT') def test_cvxopt_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='CVXOPT') def test_cvxopt_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='CVXOPT') def test_cvxopt_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='CVXOPT') def test_cvxopt_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='CVXOPT') def test_cvxopt_lp_5(self) -> None: from cvxpy.reductions.solvers.kktsolver import setup_ldl_factor StandardTestLPs.test_lp_5(solver='CVXOPT', kktsolver=setup_ldl_factor) StandardTestLPs.test_lp_5(solver='CVXOPT', kktsolver='chol') def test_cvxopt_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CVXOPT') def test_cvxopt_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='CVXOPT') def test_cvxopt_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CVXOPT') def test_cvxopt_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='CVXOPT') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='CVXOPT') def test_cvxopt_sdp_1(self) -> None: # minimization StandardTestSDPs.test_sdp_1min(solver='CVXOPT') # maximization StandardTestSDPs.test_sdp_1max(solver='CVXOPT') def test_cvxopt_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver='CVXOPT') @unittest.skipUnless('SDPA' in INSTALLED_SOLVERS, 'SDPA is not installed.') class TestSDPA(BaseTest): def test_sdpa_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='SDPA') def test_sdpa_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='SDPA') def test_sdpa_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='SDPA') def test_sdpa_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='SDPA') def test_sdpa_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='SDPA') def test_sdpa_lp_5(self) -> None: # this also tests the ability to pass solver options StandardTestLPs.test_lp_5(solver='SDPA', betaBar=0.1, gammaStar=0.8, epsilonDash=8.0E-6) def test_sdpa_lp_7(self) -> None: StandardTestLPs.test_lp_7(solver='SDPA') def test_sdpa_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='SDPA') def test_sdpa_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='SDPA') def test_sdpa_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='SDPA') def test_sdpa_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='SDPA') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='SDPA') def test_sdpa_sdp_1(self) -> None: # minimization StandardTestSDPs.test_sdp_1min(solver='SDPA') # maximization StandardTestSDPs.test_sdp_1max(solver='SDPA') def test_sdpa_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver='SDPA') def fflush() -> None: """ C code in some solvers uses libc buffering; if we want to capture log output from those solvers to use in tests, we must flush the libc buffers before trying to read the log contents from python. https://github.com/pytest-dev/pytest/issues/8753 """ import ctypes libc = ctypes.CDLL(None) libc.fflush(None) # We can't inherit from unittest.TestCase since we access some advanced pytest features. # As a result, we use the pytest skipif decorator instead of unittest.skipUnless. @pytest.mark.skipif('CBC' not in INSTALLED_SOLVERS, reason='CBC is not installed.') class TestCBC: def _cylp_checks_isProvenInfeasible(): try: # https://github.com/coin-or/CyLP/pull/150 from cylp.cy.CyCbcModel import problemStatus return problemStatus[0] == 'search completed' except ImportError: return False def test_cbc_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='CBC', duals=False) def test_cbc_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='CBC', duals=False) def test_cbc_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='CBC', duals=False) def test_cbc_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='CBC') def test_cbc_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='CBC') def test_cbc_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='CBC', duals=False) def test_cbc_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='CBC') def test_cbc_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='CBC') def test_cbc_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='CBC') def test_cbc_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='CBC') @pytest.mark.skipif(not _cylp_checks_isProvenInfeasible(), reason='CyLP <= 0.91.4 has no working integer infeasibility detection') def test_cbc_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='CBC') @pytest.mark.parametrize( "opts", [ pytest.param(opts, id=next(iter(opts.keys()))) for opts in [ {"dualTolerance": 1.0}, {"primalTolerance": 1.0}, {"maxNumIteration": 1}, {"scaling": 0}, # {"automaticScaling": True}, # Doesn't work # {"infeasibilityCost": 0.000001}, # Doesn't work {"optimizationDirection": "max"}, {"presolve": "off"}, ] ] ) def test_cbc_lp_options(self, opts: dict, capfd: pytest.LogCaptureFixture) -> None: """ Validate that cylp is actually using each option. Tentative approach: run model with verbose output with or without the specified option; verbose output should be different each way. """ # start by making sure capture buffer is empty to ensure valid results fflush() capfd.readouterr() # run the solver with verbose logging without this option and capture output sth = sths.lp_4() sth.solve(solver='CBC', logLevel=2) fflush() base = capfd.readouterr() # run the solver with verbose logging *with* the option under test try: sth.solve(solver='CBC', logLevel=2, **opts) except Exception: # if setting the option caused the case to fail, that's a pass pass else: # if the case still passes, we at least look for change in the log outputs fflush() with_opt = capfd.readouterr() assert base != with_opt def test_cbc_lp_logging(self, capfd: pytest.LogCaptureFixture) -> None: """Validate that logLevel parameter is passed to solver""" # start by making sure capture buffer is empty to ensure valid results fflush() capfd.readouterr() # for linear problems StandardTestLPs.test_lp_0(solver='CBC', duals=False, logLevel=0) fflush() quiet_output = capfd.readouterr() StandardTestLPs.test_lp_0(solver='CBC', duals=False, logLevel=5) fflush() verbose_output = capfd.readouterr() assert len(verbose_output.out) > len(quiet_output.out) # for mixed integer problems StandardTestLPs.test_mi_lp_0(solver='CBC', logLevel=0) fflush() quiet_output = capfd.readouterr() StandardTestLPs.test_mi_lp_0(solver='CBC', logLevel=5) fflush() verbose_output = capfd.readouterr() assert len(verbose_output.out) > len(quiet_output.out) @unittest.skipUnless('GLPK' in INSTALLED_SOLVERS, 'GLPK is not installed.') class TestGLPK(unittest.TestCase): def test_glpk_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='GLPK') def test_glpk_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='GLPK') def test_glpk_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='GLPK') def test_glpk_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='GLPK') def test_glpk_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='GLPK') def test_glpk_lk_5(self) -> None: StandardTestLPs.test_lp_5(solver='GLPK') def test_glpk_lp_6(self) -> None: StandardTestLPs.test_lp_6(solver='GLPK') def test_glpk_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='GLPK_MI') def test_glpk_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='GLPK_MI') def test_glpk_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='GLPK_MI') def test_glpk_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='GLPK_MI') def test_glpk_mi_lp_4(self) -> None: StandardTestLPs.test_mi_lp_4(solver='GLPK_MI') def test_glpk_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='GLPK_MI') def test_glpk_options(self) -> None: sth = sths.lp_1() import cvxopt assert "tm_lim" not in cvxopt.glpk.options sth.solve(solver='GLPK', tm_lim=100) assert "tm_lim" not in cvxopt.glpk.options sth.verify_objective(places=4) sth.check_primal_feasibility(places=4) sth.check_complementarity(places=4) sth.check_dual_domains(places=4) def test_glpk_mi_options(self) -> None: sth = sths.mi_lp_1() import cvxopt assert "tm_lim" not in cvxopt.glpk.options sth.solve(solver='GLPK_MI', tm_lim=100, verbose=True) assert "tm_lim" not in cvxopt.glpk.options sth.verify_objective(places=4) sth.verify_primal_values(places=4) @unittest.skipUnless('GLOP' in INSTALLED_SOLVERS, 'GLOP is not installed.') class TestGLOP(unittest.TestCase): def test_glop_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='GLOP') def test_glop_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='GLOP') def test_glop_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='GLOP') def test_glop_lp_3_no_preprocessing(self) -> None: from ortools.glop import parameters_pb2 params = parameters_pb2.GlopParameters() params.use_preprocessing = False StandardTestLPs.test_lp_3(solver='GLOP', parameters_proto=params) # With preprocessing enabled, Glop internally detects # INFEASIBLE_OR_UNBOUNDED. This status is translated to # MPSOLVER_INFEASIBLE. See # https://github.com/google/or-tools/blob/b37d9c786b69128f3505f15beca09e89bf078a89/ortools/linear_solver/glop_utils.cc#L25-L38. @unittest.skip('Known limitation of the GLOP interface.') def test_glop_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='GLOP') def test_glop_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='GLOP') def test_glop_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='GLOP') def test_glop_lp_6_no_preprocessing(self) -> None: from ortools.glop import parameters_pb2 params = parameters_pb2.GlopParameters() params.use_preprocessing = False StandardTestLPs.test_lp_6(solver='GLOP', parameters_proto=params) # Same issue as with test_glop_lp_3. @unittest.skip('Known limitation of the GLOP interface.') def test_glop_lp_6(self) -> None: StandardTestLPs.test_lp_6(solver='GLOP') def test_glop_bad_parameters(self) -> None: x = cp.Variable(1) prob = cp.Problem(cp.Maximize(x), [x <= 1]) with self.assertRaises(cp.error.SolverError): prob.solve(solver='GLOP', parameters_proto="not a proto") def test_glop_time_limit(self) -> None: sth = sths.lp_1() # Checks that the option doesn't error. A better test would be to solve # a large instance and check that the time limit is hit. sth.solve(solver='GLOP', time_limit_sec=1.0) @unittest.skipUnless('PDLP' in INSTALLED_SOLVERS, 'PDLP is not installed.') class TestPDLP(unittest.TestCase): def test_pdlp_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='PDLP') def test_pdlp_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='PDLP') def test_pdlp_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='PDLP') def test_pdlp_lp_3(self) -> None: sth = sths.lp_3() with self.assertWarns(Warning): sth.prob.solve(solver='PDLP') self.assertEqual(sth.prob.status, cp.settings.INFEASIBLE_OR_UNBOUNDED) # We get the precise status when presolve is disabled. def test_pdlp_lp_3_no_presolve(self) -> None: from ortools.pdlp import solvers_pb2 params = solvers_pb2.PrimalDualHybridGradientParams() params.presolve_options.use_glop = False StandardTestLPs.test_lp_3(solver='PDLP', parameters_proto=params) def test_pdlp_lp_4(self) -> None: sth = sths.lp_4() with self.assertWarns(Warning): sth.prob.solve(solver='PDLP') self.assertEqual(sth.prob.status, cp.settings.INFEASIBLE_OR_UNBOUNDED) def test_pdlp_lp_4_no_presolve(self) -> None: from ortools.pdlp import solvers_pb2 params = solvers_pb2.PrimalDualHybridGradientParams() params.presolve_options.use_glop = False StandardTestLPs.test_lp_4(solver='PDLP', parameters_proto=params) def test_pdlp_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='PDLP') def test_pdlp_lp_6(self) -> None: sth = sths.lp_6() with self.assertWarns(Warning): sth.prob.solve(solver='PDLP') self.assertEqual(sth.prob.status, cp.settings.INFEASIBLE_OR_UNBOUNDED) def test_pdlp_lp_6_no_presolve(self) -> None: from ortools.pdlp import solvers_pb2 params = solvers_pb2.PrimalDualHybridGradientParams() params.presolve_options.use_glop = False StandardTestLPs.test_lp_6(solver='PDLP', parameters_proto=params) def test_pdlp_bad_parameters(self) -> None: x = cp.Variable(1) prob = cp.Problem(cp.Maximize(x), [x <= 1]) with self.assertRaises(cp.error.SolverError): prob.solve(solver='PDLP', parameters_proto="not a proto") def test_pdlp_time_limit(self) -> None: sth = sths.lp_1() # Checks that the option doesn't error. A better test would be to solve # a large instance and check that the time limit is hit. sth.solve(solver='PDLP', time_limit_sec=1.0) @unittest.skipUnless('QOCO' in INSTALLED_SOLVERS, 'QOCO is not installed.') class TestQOCO(BaseTest): """ Unit tests for QOCO. """ def setUp(self) -> None: self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_qoco_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='QOCO') def test_qoco_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='QOCO') def test_qoco_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='QOCO') # lp 3 and 4 are skipped since QOCO cannot detect infeasibility or unboundedness. def test_qoco_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='QOCO') def test_qoco_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver='QOCO') def test_qoco_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='QOCO') def test_qoco_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='QOCO') def test_qoco_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='QOCO') def test_qoco_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='QOCO') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='QOCO') @unittest.skipUnless('CPLEX' in INSTALLED_SOLVERS, 'CPLEX is not installed.') class TestCPLEX(BaseTest): """ Unit tests for solver specific behavior. """ def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_cplex_warm_start(self) -> None: """Make sure that warm starting CPLEX behaves as expected Note: This only checks output, not whether or not CPLEX is warm starting internally """ if cp.CPLEX in INSTALLED_SOLVERS: A = cp.Parameter((2, 2)) b = cp.Parameter(2) h = cp.Parameter(2) c = cp.Parameter(2) A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([1, 1]) objective = cp.Maximize(c[0] * self.x[0] + c[1] * self.x[1]) constraints = [self.x[0] <= h[0], self.x[1] <= h[1], A @ self.x == b] prob = cp.Problem(objective, constraints) result = prob.solve(solver=cp.CPLEX, warm_start=True) self.assertEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [1, 2]) # Change A and b from the original values A.value = np.array([[0, 0], [0, 1]]) # <----- Changed b.value = np.array([0, 1]) # <----- Changed h.value = np.array([2, 2]) c.value = np.array([1, 1]) # Without setting update_eq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.CPLEX, warm_start=True) self.assertEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [2, 1]) # Change h from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([1, 1]) # <----- Changed c.value = np.array([1, 1]) # Without setting update_ineq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.CPLEX, warm_start=True) self.assertEqual(result, 2) self.assertItemsAlmostEqual(self.x.value, [1, 1]) # Change c from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([2, 1]) # <----- Changed # Without setting update_objective = False, # the results should change to the correct answer result = prob.solve(solver=cp.CPLEX, warm_start=True) self.assertEqual(result, 4) self.assertItemsAlmostEqual(self.x.value, [1, 2]) else: with self.assertRaises(Exception) as cm: prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1)), [self.x == 0]) prob.solve(solver=cp.CPLEX, warm_start=True) self.assertEqual(str(cm.exception), "The solver %s is not installed." % cp.CPLEX) def test_cplex_params(self) -> None: if cp.CPLEX in INSTALLED_SOLVERS: n, m = 10, 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) invalid_cplex_params = { "bogus": "foo" } with self.assertRaises(ValueError): problem.solve(solver=cp.CPLEX, cplex_params=invalid_cplex_params) with self.assertRaises(ValueError): problem.solve(solver=cp.CPLEX, invalid_kwarg=None) cplex_params = { "advance": 0, # int param "simplex.limits.iterations": 1000, # long param "timelimit": 1000.0, # double param "workdir": '"mydir"', # string param } problem.solve(solver=cp.CPLEX, cplex_params=cplex_params) def test_cplex_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='CPLEX') def test_cplex_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='CPLEX') def test_cplex_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='CPLEX') def test_cplex_lp_3(self) -> None: # CPLEX initially produces an INFEASIBLE_OR_UNBOUNDED status sth = sths.lp_3() with self.assertWarns(Warning): sth.prob.solve(solver='CPLEX') self.assertEqual(sth.prob.status, cp.settings.INFEASIBLE_OR_UNBOUNDED) # Determine the precise status with reoptimize=True. StandardTestLPs.test_lp_3(solver='CPLEX', reoptimize=True) def test_cplex_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='CPLEX') def test_cplex_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='CPLEX') def test_cplex_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='CPLEX') def test_cplex_socp_1(self) -> None: # Parameters are set due to a minor dual-variable related # presolve bug in CPLEX, which will be fixed in the next # CPLEX release. StandardTestSOCPs.test_socp_1(solver='CPLEX', places=2, cplex_params={ "preprocessing.presolve": 0, "preprocessing.reduce": 2}) def test_cplex_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='CPLEX') def test_cplex_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='CPLEX') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='CPLEX') def test_cplex_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='CPLEX') def test_cplex_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='CPLEX') def test_cplex_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='CPLEX') def test_cplex_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='CPLEX') def test_cplex_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='CPLEX') def test_cplex_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='CPLEX', places=3) def test_cplex_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver='CPLEX') @unittest.skipUnless('GUROBI' in INSTALLED_SOLVERS, 'GUROBI is not installed.') class TestGUROBI(BaseTest): """NOTE: solves of LPs (or MILPs) get routed through GUROBI's QP interface! So many of these tests are testing the behavior of qurobi_qpif.py""" def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_gurobi_warm_start(self) -> None: """Make sure that warm starting Gurobi behaves as expected Note: This only checks output, not whether or not Gurobi is warm starting internally """ if cp.GUROBI in INSTALLED_SOLVERS: import gurobipy import numpy as np A = cp.Parameter((2, 2)) b = cp.Parameter(2) h = cp.Parameter(2) c = cp.Parameter(2) A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([1, 1]) objective = cp.Maximize(c[0] * self.x[0] + c[1] * self.x[1]) constraints = [self.x[0]**2 <= h[0]**2, self.x[1] <= h[1], A @ self.x == b] prob = cp.Problem(objective, constraints) result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertAlmostEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [1, 2]) # Change A and b from the original values A.value = np.array([[0, 0], [0, 1]]) # <----- Changed b.value = np.array([0, 1]) # <----- Changed h.value = np.array([2, 2]) c.value = np.array([1, 1]) # Without setting update_eq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertAlmostEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [2, 1]) # Change h from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([1, 1]) # <----- Changed c.value = np.array([1, 1]) # Without setting update_ineq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertAlmostEqual(result, 2) self.assertItemsAlmostEqual(self.x.value, [1, 1]) # Change c from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([2, 1]) # <----- Changed # Without setting update_objective = False, # the results should change to the correct answer result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertEqual(result, 4) self.assertItemsAlmostEqual(self.x.value, [1, 2]) # Try creating a new problem and setting x.value. init_value = np.array([2, 3]) self.x.value = init_value prob = cp.Problem(objective, constraints) result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertEqual(result, 4) self.assertItemsAlmostEqual(self.x.value, [1, 2]) # Check that "start" value was set appropriately. model = prob.solver_stats.extra_stats model_x = model.getVars() for i in range(self.x.size): assert init_value[i] == model_x[i].start assert np.isclose(self.x.value[i], model_x[i].x) # Test with matrix variable. z = cp.Variable() Y = cp.Variable((3, 2)) Y_val = np.reshape(np.arange(6), (3, 2)) Y.value = Y_val + 1 objective = cp.Maximize(z + cp.sum(Y)) constraints = [Y <= Y_val, z <= 2] prob = cp.Problem(objective, constraints) result = prob.solve(solver=cp.GUROBI, warm_start=True) self.assertEqual(result, Y_val.sum() + 2) self.assertAlmostEqual(z.value, 2) self.assertItemsAlmostEqual(Y.value, Y_val) # Check that "start" value was set appropriately. model = prob.solver_stats.extra_stats model_x = model.getVars() assert gurobipy.GRB.UNDEFINED == model_x[0].start assert np.isclose(2, model_x[0].x) for i in range(1, Y.size + 1): row = (i - 1) % Y.shape[0] col = (i - 1) // Y.shape[0] assert Y_val[row, col] + 1 == model_x[i].start assert np.isclose(Y.value[row, col], model_x[i].x) else: with self.assertRaises(Exception) as cm: prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1)), [self.x == 0]) prob.solve(solver=cp.GUROBI, warm_start=True) self.assertEqual(str(cm.exception), "The solver %s is not installed." % cp.GUROBI) def test_gurobi_time_limit_no_solution(self) -> None: """Make sure that if Gurobi terminates due to a time limit before finding a solution: 1) no error is raised, 2) solver stats are returned. The test is skipped if something changes on Gurobi's side so that: - a solution is found despite a time limit of zero, - a different termination criteria is hit first. """ if cp.GUROBI in INSTALLED_SOLVERS: import gurobipy objective = cp.Minimize(self.x[0]) constraints = [cp.square(self.x[0]) <= 1] prob = cp.Problem(objective, constraints) try: prob.solve(solver=cp.GUROBI, TimeLimit=0.0) except Exception as e: self.fail("An exception %s is raised instead of returning a result." % e) extra_stats = None solver_stats = getattr(prob, "solver_stats", None) if solver_stats: extra_stats = getattr(solver_stats, "extra_stats", None) self.assertTrue(extra_stats, "Solver stats have not been returned.") nb_solutions = getattr(extra_stats, "SolCount", None) if nb_solutions: self.skipTest("Gurobi has found a solution, the test is not relevant anymore.") solver_status = getattr(extra_stats, "Status", None) if solver_status != gurobipy.GRB.TIME_LIMIT: self.skipTest("Gurobi terminated for a different reason than reaching time limit, " "the test is not relevant anymore.") else: with self.assertRaises(Exception) as cm: prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1)), [self.x == 0]) prob.solve(solver=cp.GUROBI, TimeLimit=0) self.assertEqual(str(cm.exception), "The solver %s is not installed." % cp.GUROBI) def test_gurobi_environment(self) -> None: """Tests that Gurobi environments can be passed to Model. Gurobi environments can include licensing and model parameter data. """ if cp.GUROBI in INSTALLED_SOLVERS: import gurobipy # Set a few parameters to random values close to their defaults params = { 'MIPGap': np.random.random(), # range {0, INFINITY} 'AggFill': np.random.randint(10), # range {-1, MAXINT} 'PerturbValue': np.random.random(), # range: {0, INFINITY} } # Create a custom environment and set some parameters custom_env = gurobipy.Env() for k, v in params.items(): custom_env.setParam(k, v) # Testing Conic Solver Interface sth = StandardTestSOCPs.test_socp_0(solver='GUROBI', env=custom_env) model = sth.prob.solver_stats.extra_stats for k, v in params.items(): # https://www.gurobi.com/documentation/9.1/refman/py_model_getparaminfo.html name, p_type, p_val, p_min, p_max, p_def = model.getParamInfo(k) self.assertEqual(v, p_val) else: with self.assertRaises(Exception) as cm: prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1)), [self.x == 0]) prob.solve(solver=cp.GUROBI, TimeLimit=0) self.assertEqual(str(cm.exception), "The solver %s is not installed." % cp.GUROBI) def test_gurobi_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='GUROBI') def test_gurobi_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='GUROBI') def test_gurobi_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='GUROBI') def test_gurobi_lp_3(self) -> None: # GUROBI initially produces an INFEASIBLE_OR_UNBOUNDED status sth = sths.lp_3() with self.assertWarns(Warning): sth.prob.solve(solver='GUROBI') self.assertEqual(sth.prob.status, cp.settings.INFEASIBLE_OR_UNBOUNDED) # The user disables presolve and so makes reoptimization unnecessary StandardTestLPs.test_lp_3(solver='GUROBI', InfUnbdInfo=1) # The user determines the precise status with reoptimize=True StandardTestLPs.test_lp_3(solver='GUROBI', reoptimize=True) def test_gurobi_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='GUROBI', reoptimize=True) def test_gurobi_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='GUROBI') def test_gurobi_lp_bound_attr(self) -> None: sth = StandardTestLPs.test_lp_bound_attr(solver='GUROBI') # check that the bounds do reach the solver and don't just generate constraints model = sth.prob.solver_stats.extra_stats expected_bounds = sth.prob.variables()[0].bounds actual_lb = [v.LB for v in model.getVars()] actual_ub = [v.UB for v in model.getVars()] np.testing.assert_equal([actual_lb, actual_ub], expected_bounds) def test_gurobi_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='GUROBI') def test_gurobi_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='GUROBI') def test_gurobi_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='GUROBI') def test_gurobi_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='GUROBI') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='GUROBI') def test_gurobi_socp_bound_attr(self) -> None: sth = StandardTestSOCPs.test_socp_bounds_attr(solver='GUROBI') # check that the bounds do reach the solver and don't just generate constraints model = sth.prob.solver_stats.extra_stats x = model.getVarByName("x_0") expected_bounds = sth.prob.variables()[0].bounds actual_bounds = [x.LB, x.UB] np.testing.assert_equal(actual_bounds, expected_bounds) def test_gurobi_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='GUROBI') def test_gurobi_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='GUROBI') def test_gurobi_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='GUROBI') def test_gurobi_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='GUROBI') def test_gurobi_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='GUROBI') def test_gurobi_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='GUROBI', places=2) def test_gurobi_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver='GUROBI') @unittest.skipUnless('XPRESS' in INSTALLED_SOLVERS, 'XPRESS is not installed.') class TestXPRESS(BaseTest): def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x', integer=True) self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_xpress_warm_start(self) -> None: """Make sure that warm starting Xpress behaves as expected Note: Xpress does not have warmstart yet, it will re-solve problem from scratch """ if cp.XPRESS in INSTALLED_SOLVERS: import numpy as np A = cp.Parameter((2, 2)) b = cp.Parameter(2) h = cp.Parameter(2) c = cp.Parameter(2) A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([1, 1]) objective = cp.Maximize(c[0] * self.x[0] + c[1] * self.x[1]) constraints = [self.x[0] <= h[0], self.x[1] <= h[1], A @ self.x == b] prob = cp.Problem(objective, constraints) result = prob.solve(solver=cp.XPRESS, warm_start=True) self.assertEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [1, 2]) # Change A and b from the original values A.value = np.array([[0, 0], [0, 1]]) # <----- Changed b.value = np.array([0, 1]) # <----- Changed h.value = np.array([2, 2]) c.value = np.array([1, 1]) # Without setting update_eq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.XPRESS, warm_start=True) self.assertEqual(result, 3) self.assertItemsAlmostEqual(self.x.value, [2, 1]) # Change h from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([1, 1]) # <----- Changed c.value = np.array([1, 1]) # Without setting update_ineq_constrs = False, # the results should change to the correct answer result = prob.solve(solver=cp.XPRESS, warm_start=True) self.assertEqual(result, 2) self.assertItemsAlmostEqual(self.x.value, [1, 1]) # Change c from the original values A.value = np.array([[1, 0], [0, 0]]) b.value = np.array([1, 0]) h.value = np.array([2, 2]) c.value = np.array([2, 1]) # <----- Changed # Without setting update_objective = False, # the results should change to the correct answer result = prob.solve(solver=cp.XPRESS, warm_start=True) self.assertEqual(result, 4) self.assertItemsAlmostEqual(self.x.value, [1, 2]) else: with self.assertRaises(Exception) as cm: prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1)), [self.x == 0]) prob.solve(solver=cp.XPRESS, warm_start=True) self.assertEqual(str(cm.exception), "The solver %s is not installed." % cp.XPRESS) def test_xpress_params(self) -> None: if cp.XPRESS in INSTALLED_SOLVERS: n, m = 10, 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) params = { "lpiterlimit": 1000, # maximum number of simplex iterations "maxtime": 1000 # time limit } problem.solve(solver=cp.XPRESS, **params) def test_xpress_iis_none(self) -> None: if cp.XPRESS in INSTALLED_SOLVERS: A = np.array([[2, 1], [1, 2], [-3, -3]]) b = np.array([2, 2, -5]) x = cp.Variable(2) objective = cp.Minimize(cp.norm2(x)) constraint = [A @ x <= b] problem = cp.Problem(objective, constraint) params = {'save_iis': 0} problem.solve(solver=cp.XPRESS, **params) def test_xpress_iis_full(self) -> None: if cp.XPRESS in INSTALLED_SOLVERS: A = np.array([[2, 1], [1, 2], [-3, -3]]) b = np.array([2, 2, -5]) x = cp.Variable(2) objective = cp.Minimize(cp.norm2(x)) constraint = [A @ x <= b] problem = cp.Problem(objective, constraint) params = {'save_iis': -1} problem.solve(solver=cp.XPRESS, **params) assert 'XPRESS_IIS' in problem.solution.attr def test_xpress_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='XPRESS') def test_xpress_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='XPRESS') def test_xpress_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='XPRESS') def test_xpress_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='XPRESS') def test_xpress_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='XPRESS') def test_xpress_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='XPRESS') def test_xpress_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='XPRESS') def test_xpress_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='XPRESS') def test_xpress_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='XPRESS') def test_xpress_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='XPRESS') def test_xpress_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='XPRESS') def test_xpress_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='XPRESS') def test_xpress_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='XPRESS') def test_xpress_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='XPRESS') def test_xpress_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver='XPRESS') @unittest.skipUnless('NAG' in INSTALLED_SOLVERS, 'NAG is not installed.') class TestNAG(BaseTest): def test_nag_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='NAG') def test_nag_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='NAG') def test_nag_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='NAG') def test_nag_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='NAG') def test_nag_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='NAG') def test_nag_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='NAG') def test_nag_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='NAG') def test_nag_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='NAG') def test_nag_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='NAG') def test_nag_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='NAG') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='NAG') def test_nag_quad_obj(self) -> None: """Test NAG canonicalization with a quadratic objective. """ x = cp.Variable(2) expr = cp.sum_squares(x) constr = [x >= 1] prob = cp.Problem(cp.Minimize(expr), constr) data = prob.get_problem_data(solver='NAG') self.assertItemsAlmostEqual(data[0]["P"].A, 2*np.eye(2)) solution1 = prob.solve(solver='NAG') # When use_quad_obj = False, the quadratic objective is # canonicalized to a SOC constraint. prob = cp.Problem(cp.Minimize(expr), constr) solver_opts = {"use_quad_obj": False} data = prob.get_problem_data(solver='NAG', solver_opts=solver_opts) assert "P" not in data[0] solution2 = prob.solve(solver='NAG', **solver_opts) assert np.isclose(solution1, solution2) # Check that there is no P for non-quadratic objectives. expr = cp.norm(x, 1) prob = cp.Problem(cp.Minimize(expr), constr) data = prob.get_problem_data(solver='NAG') assert "P" not in data[0] @unittest.skipUnless("SCIP" in INSTALLED_SOLVERS, "SCIP is not installed.") class TestSCIP(unittest.TestCase): def test_scip_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver="SCIP") def test_scip_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver="SCIP") def test_scip_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver="SCIP", duals=False) def test_scip_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver="SCIP") def test_scip_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver="SCIP") def test_scip_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver="SCIP") def test_scip_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver="SCIP", places=2, duals=False) def test_scip_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver="SCIP", places=2, duals=False) def test_scip_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver="SCIP", duals=False) # axis 1 StandardTestSOCPs.test_socp_3ax1(solver="SCIP", duals=False) def test_scip_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver="SCIP") def test_scip_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver="SCIP") def test_scip_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver="SCIP") def test_scip_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver="SCIP") def test_scip_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver="SCIP") def test_scip_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver="SCIP", places=3) def test_scip_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver="SCIP") def get_simple_problem(self): """Example problem that can be used within additional tests.""" x = cp.Variable() y = cp.Variable() constraints = [ x >= 0, # x must be positive y >= 1, # y must be greater or equal to 1 x + y <= 4 ] obj = cp.Maximize(x) prob = cp.Problem(obj, constraints) return prob def test_scip_test_params__no_params_set(self) -> None: prob = self.get_simple_problem() prob.solve(solver="SCIP") # Important that passes without raising an error also check obj. assert prob.value == 3 def test_scip_test_params__valid_params(self) -> None: prob = self.get_simple_problem() prob.solve(solver="SCIP", gp=False) # Important that passes without raising an error also check obj. assert prob.value == 3 def test_scip_test_params__valid_scip_params(self) -> None: prob = self.get_simple_problem() prob.solve(solver="SCIP", scip_params={"lp/fastmip": 1, "limits/gap": 0.1}) # Important that passes without raising an error also check obj. assert prob.value == 3 def test_scip_test_params__invalid_params(self) -> None: prob = self.get_simple_problem() # Since an invalid NON-scip param is passed, an error is expected # to be raised when calling solve. with pytest.raises(KeyError) as ke: prob.solve(solver="SCIP", a="what?") exc = "One or more solver params in ['a'] are not valid: 'Not a valid parameter name'" assert ke.exception == exc def test_scip_test_params__invalid_scip_params(self) -> None: prob = self.get_simple_problem() # Since an invalid SCIP param is passed, an error is expected # to be raised when calling solve. with pytest.raises(KeyError) as ke: prob.solve(solver="SCIP", scip_params={"a": "what?"}) exc = "One or more scip params in ['a'] are not valid: 'Not a valid parameter name'" assert ke.exception == exc def test_scip_time_limit_reached(self) -> None: sth = sths.mi_lp_7() # TODO doesn't work on windows. # run without enough time to find optimum # sth.solve(solver="SCIP", scip_params={"limits/time": 0.01}) # assert sth.prob.status == cp.OPTIMAL_INACCURATE # assert all([v.value is not None for v in sth.prob.variables()]) # run without enough time to do anything with pytest.raises(cp.error.SolverError) as se: sth.solve(solver="SCIP", scip_params={"limits/time": 0.0}) exc = "Solver 'SCIP' failed. " \ "Try another solver, or solve with verbose=True for more information." assert str(se.value) == exc # We can't inherit from unittest.TestCase since we access some advanced pytest features. # As a result, we use the pytest skipif decorator instead of unittest.skipUnless. @pytest.mark.skipif('HIGHS' not in INSTALLED_SOLVERS, reason="HiGHS is not installed.") class TestHIGHS: @pytest.mark.parametrize( "problem", [ StandardTestLPs.test_lp_0, StandardTestLPs.test_lp_1, StandardTestLPs.test_lp_2, StandardTestLPs.test_lp_3, StandardTestLPs.test_lp_4, StandardTestLPs.test_lp_5, StandardTestLPs.test_mi_lp_0, StandardTestLPs.test_mi_lp_1, StandardTestLPs.test_mi_lp_2, StandardTestLPs.test_mi_lp_3, StandardTestLPs.test_mi_lp_4, StandardTestLPs.test_mi_lp_5, ], ) def test_highs_solving(self, problem) -> None: # HACK needed to use the HiGHS conic interface rather than # the QP interface for LPs. from cvxpy.reductions.solvers.conic_solvers.highs_conif import HIGHS solver = HIGHS() solver.name = lambda: "HIGHS CONIC" problem(solver=solver) @pytest.mark.parametrize( ["problem", "confirmation_string"], [ (StandardTestLPs.test_lp_2, "Solving LP with useful basis"), (StandardTestLPs.test_mi_lp_2, "Assessing feasibility of MIP"), ], ) def test_highs_warm_start(self, problem, confirmation_string, capfd) -> None: """Test that HiGHS actually gets warm started for each problem type. Args: problem: LP or MIP. confirmation_string: Regex that confirms that HiGHS was warm started. capfd: Captures stdout to search for confirmation_string. """ p = problem(cp.HIGHS, verbose=True) p.prob.solve(cp.HIGHS, warm_start=True, verbose=True) captured = capfd.readouterr() assert re.search(confirmation_string, captured.out) is not None def test_highs_nonstandard_name(self) -> None: """Test HiGHS solver with non-capitalized solver name.""" # https://github.com/cvxpy/cvxpy/issues/2751 StandardTestLPs.test_lp_0(solver="HiGHS") @pytest.mark.parametrize( "problem", # it is enough to validate the options with one problem from each type [StandardTestLPs.test_lp_0, StandardTestLPs.test_mi_lp_0], ) def test_highs_options(self, problem) -> None: with pytest.raises(ValueError): problem(solver=cp.HIGHS, highs_options={"invalid_highs_option": None}) with pytest.raises(ValueError): problem(solver=cp.HIGHS, invalid_highs_option=None) invalid_value = "_invalid_value_" with pytest.raises(ValueError): problem(solver=cp.HIGHS, highs_options={"threads": invalid_value}) with pytest.raises(ValueError): problem(solver=cp.HIGHS, threads=invalid_value) # Duplicate options with pytest.raises(TypeError): problem(solver=cp.HIGHS, presolve="off", highs_options=dict(presolve="off")) # Happy path highs_options = { "solver": "simplex", # string option -- also `"solver"` shouldn't name-clash "output_flag": True, # bool option "time_limit": 0.123, # double option "random_seed": 1234, # int option # no need to optimize for real "mip_rel_gap": 1.0, "primal_feasibility_tolerance": 1e-3, "dual_feasibility_tolerance": 1e-3, # advanced option, not available on HighsOptions "presolve_rule_off": 2**10, } problem(solver=cp.HIGHS, highs_options=highs_options) class TestAllSolvers(BaseTest): def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_installed_solvers(self) -> None: """Test the list of installed solvers. """ from cvxpy.reductions.solvers.defines import ( INSTALLED_SOLVERS, SOLVER_MAP_CONIC, SOLVER_MAP_QP, ) prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1) + 1.0), [self.x == 0]) for solver in SOLVER_MAP_CONIC.keys(): if solver in INSTALLED_SOLVERS: prob.solve(solver=solver) self.assertAlmostEqual(prob.value, 1.0) self.assertItemsAlmostEqual(self.x.value, [0, 0]) else: with self.assertRaises(Exception) as cm: prob.solve(solver=solver) self.assertEqual(str(cm.exception), "The solver %s is not installed." % solver) for solver in SOLVER_MAP_QP.keys(): if solver in INSTALLED_SOLVERS: prob.solve(solver=solver) self.assertItemsAlmostEqual(self.x.value, [0, 0]) else: with self.assertRaises(Exception) as cm: prob.solve(solver=solver) self.assertEqual(str(cm.exception), "The solver %s is not installed." % solver) def test_mixed_integer_behavior(self) -> None: x = cp.Variable(2, name='x', integer=True) objective = cp.Minimize(cp.sum(x)) prob = cp.Problem(objective, [x >= 0]) if INSTALLED_MI_SOLVERS == [cp.ECOS_BB]: with pytest.raises(cp.error.SolverError, match="You need a mixed-integer " "solver for this model"): prob.solve() else: prob.solve() self.assertItemsAlmostEqual(x.value, [0, 0]) @unittest.skipUnless('ECOS' in INSTALLED_SOLVERS, 'ECOS_BB is not installed.') class TestECOS_BB(unittest.TestCase): def test_ecos_bb_explicit_only(self) -> None: """Test that ECOS_BB isn't chosen by default. """ x = cp.Variable(1, name='x', integer=True) objective = cp.Minimize(cp.sum(x)) prob = cp.Problem(objective, [x >= 0]) if INSTALLED_MI_SOLVERS != [cp.ECOS_BB]: prob.solve() assert prob.solver_stats.solver_name != cp.ECOS_BB else: with pytest.raises(cp.error.SolverError, match="You need a mixed-integer " "solver for this model"): prob.solve() def test_ecos_bb_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='ECOS_BB') def test_ecos_bb_lp_1(self) -> None: # default settings StandardTestLPs.test_lp_1(solver='ECOS_BB') # require a basic feasible solution StandardTestLPs.test_lp_1(solver='ECOS_BB') def test_ecos_bb_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='ECOS_BB') def test_ecos_bb_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='ECOS_BB') def test_ecos_bb_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='ECOS_BB') def test_ecos_bb_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='ECOS_BB') def test_ecos_bb_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='ECOS_BB') def test_ecos_bb_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='ECOS_BB') def test_ecos_bb_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='ECOS_BB') def test_ecos_bb_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='ECOS_BB') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='ECOS_BB') def test_ecos_bb_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='ECOS_BB') def test_ecos_bb_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='ECOS_BB') def test_ecos_bb_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='ECOS_BB') @pytest.mark.skip(reason="Known bug in ECOS BB") def test_ecos_bb_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='ECOS_BB') def test_ecos_bb_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='ECOS_BB') def test_ecos_bb_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='ECOS_BB') @pytest.mark.skip(reason="Known bug in ECOS BB") def test_ecos_bb_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='ECOS_BB') class TestSCIPY(unittest.TestCase): def setUp(self): import scipy self.d = Version(scipy.__version__) >= Version('1.7.0') def test_scipy_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='SCIPY', duals=self.d) def test_scipy_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='SCIPY', duals=self.d) def test_scipy_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='SCIPY', duals=self.d) def test_scipy_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='SCIPY') def test_scipy_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='SCIPY') def test_scipy_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='SCIPY', duals=self.d) def test_scipy_lp_solver_stats(self) -> None: sth = StandardTestLPs.test_lp_0(solver='SCIPY', duals=self.d) # Equal because presolve might directly find the solution in 0 iterations self.assertGreaterEqual(sth.prob.solver_stats.num_iters, 0) @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_4(self) -> None: StandardTestLPs.test_mi_lp_4(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='SCIPY') @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_time_limit_reached(self) -> None: sth = sths.mi_lp_7() # We only check that the option does not raise an error. sth.solve(solver='SCIPY', scipy_options={"time_limit": 0.1}) @unittest.skipUnless('SCIPY' in INSTALLED_MI_SOLVERS, 'SCIPY version cannot solve MILPs') def test_scipy_mi_solver_stats(self) -> None: sth = sths.mi_lp_7() sth.solve(solver='SCIPY') self.assertTrue("mip_gap" in sth.prob.solver_stats.extra_stats) self.assertTrue("mip_node_count" in sth.prob.solver_stats.extra_stats) self.assertTrue("mip_dual_bound" in sth.prob.solver_stats.extra_stats) @unittest.skipUnless('COPT' in INSTALLED_SOLVERS, 'COPT is not installed.') class TestCOPT(unittest.TestCase): def test_copt_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='COPT') def test_copt_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='COPT') def test_copt_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='COPT') def test_copt_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='COPT') def test_copt_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='COPT') def test_copt_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='COPT') def test_copt_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='COPT') def test_copt_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='COPT', places=3) def test_copt_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='COPT') def test_copt_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='COPT') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='COPT') def test_copt_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='COPT') def test_copt_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='COPT') def test_copt_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='COPT') def test_copt_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='COPT') def test_copt_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='COPT') def test_copt_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver='COPT') def test_copt_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='COPT') def test_copt_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver='COPT') def test_copt_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver='COPT') def test_copt_sdp_1min(self) -> None: StandardTestSDPs.test_sdp_1min(solver='COPT') def test_copt_sdp_1max(self) -> None: StandardTestSDPs.test_sdp_1max(solver='COPT') def test_copt_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver='COPT') def test_copt_params(self) -> None: n = 10 m = 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) with self.assertRaises(AttributeError): problem.solve(solver=cp.COPT, invalid_kwarg=None) # Valid arg. problem.solve(solver=cp.COPT, feastol=1e-9) @unittest.skipUnless('COSMO' in INSTALLED_SOLVERS, 'COSMO is not installed.') class TestCOSMO(BaseTest): """Unit tests for COSMO solver interface.""" def setUp(self) -> None: self.a = cp.Variable(name='a') self.b = cp.Variable(name='b') self.c = cp.Variable(name='c') self.x = cp.Variable(2, name='x') self.y = cp.Variable(3, name='y') self.z = cp.Variable(2, name='z') self.A = cp.Variable((2, 2), name='A') self.B = cp.Variable((2, 2), name='B') self.C = cp.Variable((3, 2), name='C') def test_cosmo_options(self) -> None: """Test that COSMO solver options work.""" # Test basic options with simple problem prob = cp.Problem(cp.Minimize(cp.norm(self.x, 1) + 1.0), [self.x == 0]) # Test with default settings prob.solve(solver=cp.COSMO, verbose=True) self.assertAlmostEqual(prob.value, 1.0) self.assertItemsAlmostEqual(self.x.value, [0, 0]) # Test with custom settings prob.solve(solver=cp.COSMO, eps_abs=1e-4, verbose=True) self.assertAlmostEqual(prob.value, 1.0) self.assertItemsAlmostEqual(self.x.value, [0, 0]) def test_cosmo_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver='COSMO') def test_cosmo_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver='COSMO') def test_cosmo_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver='COSMO') def test_cosmo_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver='COSMO') def test_cosmo_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver='COSMO') def test_cosmo_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='COSMO') def test_cosmo_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver='COSMO') def test_cosmo_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver='COSMO') def test_cosmo_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver='COSMO') def test_cosmo_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver='COSMO') def test_cosmo_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver='COSMO') # axis 1 StandardTestSOCPs.test_socp_3ax1(solver='COSMO') def test_cosmo_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver='COSMO') def test_cosmo_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver='COSMO') def test_cosmo_pcp_1(self) -> None: StandardTestPCPs.test_pcp_1(solver='COSMO') def test_cosmo_pcp_2(self) -> None: StandardTestPCPs.test_pcp_2(solver='COSMO') def test_cosmo_pcp_3(self) -> None: StandardTestPCPs.test_pcp_3(solver='COSMO') def test_cosmo_sdp_1min(self) -> None: StandardTestSDPs.test_sdp_1min(solver='COSMO') def test_cosmo_sdp_2(self) -> None: places = 3 sth = sths.sdp_2() sth.solve(solver='COSMO') sth.verify_objective(places) sth.check_primal_feasibility(places) sth.verify_primal_values(places) sth.check_complementarity(places) sth.check_dual_domains(places) def test_warm_start(self) -> None: """Test warm starting. """ x = cp.Variable(10) obj = cp.Minimize(cp.sum(cp.exp(x))) prob = cp.Problem(obj, [cp.sum(x) == 1]) result = prob.solve(solver=cp.COSMO) time = prob.solver_stats.solve_time result2 = prob.solve(solver=cp.COSMO, warm_start=True) time2 = prob.solver_stats.solve_time self.assertAlmostEqual(result2, result, places=2) print(time > time2) def is_knitro_available(): """Check if KNITRO is installed and a license is available.""" if 'KNITRO' not in INSTALLED_SOLVERS: return False try: import knitro # type: ignore # Try to create and delete a Knitro solver instance kc = knitro.KN_new() if kc is None: return False knitro.KN_free(kc) return True except Exception: return False @unittest.skipUnless(is_knitro_available(), 'KNITRO is not installed or license is not available.') class TestKNITRO(BaseTest): def test_knitro_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver=cp.KNITRO) def test_knitro_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver=cp.KNITRO) def test_knitro_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver=cp.KNITRO) def test_knitro_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver=cp.KNITRO) def test_knitro_lp_4(self) -> None: StandardTestLPs.test_lp_4(solver=cp.KNITRO) def test_knitro_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver=cp.KNITRO) def test_knitro_lp_6(self) -> None: StandardTestLPs.test_lp_6(solver=cp.KNITRO) def test_knitro_lp_bound_attr(self) -> None: StandardTestLPs.test_lp_bound_attr(solver=cp.KNITRO) def test_knitro_socp_0(self) -> None: StandardTestSOCPs.test_socp_0(solver=cp.KNITRO) def test_knitro_socp_1(self) -> None: StandardTestSOCPs.test_socp_1(solver=cp.KNITRO) def test_knitro_socp_2(self) -> None: StandardTestSOCPs.test_socp_2(solver=cp.KNITRO) def test_knitro_socp_3(self) -> None: # axis 0 StandardTestSOCPs.test_socp_3ax0(solver=cp.KNITRO) # axis 1 StandardTestSOCPs.test_socp_3ax1(solver=cp.KNITRO) def test_knitro_socp_bounds_attr(self) -> None: StandardTestSOCPs.test_socp_bounds_attr(solver=cp.KNITRO) def test_knitro_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver=cp.KNITRO) def test_knitro_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver=cp.KNITRO) def test_knitro_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver=cp.KNITRO) def test_knitro_mi_lp_3(self) -> None: StandardTestLPs.test_mi_lp_3(solver=cp.KNITRO) def test_knitro_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver=cp.KNITRO) def test_knitro_mi_socp_1(self) -> None: StandardTestSOCPs.test_mi_socp_1(solver=cp.KNITRO) def test_knitro_mi_socp_2(self) -> None: StandardTestSOCPs.test_mi_socp_2(solver=cp.KNITRO) def test_knitro_qp_0(self) -> None: StandardTestQPs.test_qp_0(solver=cp.KNITRO) def test_knitro_expcone_1(self) -> None: StandardTestECPs.test_expcone_1(solver=cp.KNITRO) def test_knitro_pcp_1(self) -> None: StandardTestPCPs.test_pcp_1(solver=cp.KNITRO) def test_knitro_pcp_2(self) -> None: StandardTestPCPs.test_pcp_2(solver=cp.KNITRO) def test_knitro_pcp_3(self) -> None: StandardTestPCPs.test_pcp_3(solver=cp.KNITRO) def test_knitro_mi_pcp_0(self) -> None: StandardTestPCPs.test_mi_pcp_0(solver=cp.KNITRO) def test_knitro_sdp_1min(self) -> None: StandardTestSDPs.test_sdp_1min(solver=cp.KNITRO) def test_knitro_sdp_1max(self) -> None: StandardTestSDPs.test_sdp_1max(solver=cp.KNITRO) def test_knitro_sdp_2(self) -> None: StandardTestSDPs.test_sdp_2(solver=cp.KNITRO) def test_knitro_exp_soc_1(self) -> None: StandardTestMixedCPs.test_exp_soc_1(solver=cp.KNITRO) def test_knitro_sdp_pcp_1(self) -> None: StandardTestMixedCPs.test_sdp_pcp_1(solver=cp.KNITRO) def test_knitro_params(self) -> None: n = 10 m = 4 np.random.seed(0) A = np.random.randn(m, n) x = np.random.randn(n) y = A.dot(x) # Solve a simple basis pursuit problem for testing purposes. z = cp.Variable(n) objective = cp.Minimize(cp.norm1(z)) constraints = [A @ z == y] problem = cp.Problem(objective, constraints) with self.assertRaises(Exception): opts = {"a": "invalid"} problem.solve(solver=cp.KNITRO, **opts) with self.assertRaises(Exception): opts = {"algorithm": "invalid"} problem.solve(solver=cp.KNITRO, **opts) opts = {"algorithm": 0} problem.solve(solver=cp.KNITRO, **opts) @unittest.skipUnless("CUOPT" in INSTALLED_SOLVERS, "CUOPT is not installed.") class TestCUOPT(unittest.TestCase): import os kwargs={"pdlp_solver_mode": os.environ.get("CUOPT_PDLP_SOLVER_MODE", "Stable2"), "solver_method": os.environ.get("CUOPT_SOLVER_METHOD", 0) } def test_cuopt_lp_0(self) -> None: StandardTestLPs.test_lp_0(solver="CUOPT", duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_1(self) -> None: StandardTestLPs.test_lp_1(solver="CUOPT", duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_2(self) -> None: StandardTestLPs.test_lp_2(solver="CUOPT", duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_3(self) -> None: StandardTestLPs.test_lp_3(solver="CUOPT", duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_4(self) -> None: # In this case cuopt throws an exception because there are crossing # variable bounds x <= 0 and x >= 1 try: StandardTestLPs.test_lp_4(solver="CUOPT", duals=True, places=4, **TestCUOPT.kwargs) except Exception as e: assert "crossing bounds" in str(e) def test_cuopt_lp_5(self) -> None: StandardTestLPs.test_lp_5(solver='CUOPT', duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_6(self) -> None: StandardTestLPs.test_lp_5(solver='CUOPT', duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_lp_7(self) -> None: StandardTestLPs.test_lp_5(solver='CUOPT', duals=True, places=4, **TestCUOPT.kwargs) def test_cuopt_mi_lp_0(self) -> None: StandardTestLPs.test_mi_lp_0(solver='CUOPT', **TestCUOPT.kwargs) def test_cuopt_mi_lp_1(self) -> None: StandardTestLPs.test_mi_lp_1(solver='CUOPT', **TestCUOPT.kwargs) def test_cuopt_mi_lp_2(self) -> None: StandardTestLPs.test_mi_lp_2(solver='CUOPT', **TestCUOPT.kwargs) def test_cuopt_mi_lp_3(self) -> None: TestCUOPT.kwargs["time_limit"] = 5 try: StandardTestLPs.test_mi_lp_3(solver='CUOPT', **TestCUOPT.kwargs) finally: del TestCUOPT.kwargs["time_limit"] # This is an unconstrained problem, which cuopt doesn't handle. # Error message from cvxpy should be returned def test_cuopt_mi_lp_4(self) -> None: try: StandardTestLPs.test_mi_lp_4(solver='CUOPT', **TestCUOPT.kwargs) except Exception as e: assert "there are not enough constraints in the problem" in str(e) def test_cuopt_mi_lp_5(self) -> None: StandardTestLPs.test_mi_lp_5(solver='CUOPT', **TestCUOPT.kwargs, time_limit=5) def test_cuopt_mi_lp_7(self) -> None: StandardTestLPs.test_mi_lp_5(solver='CUOPT', **TestCUOPT.kwargs, time_limit=5)