""" Copyright 2013 Steven Diamond 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 warnings import numpy as np import pytest import scipy.sparse as sp from hypothesis import assume, given from hypothesis.extra.numpy import ( arrays, basic_indices, broadcastable_shapes, integer_array_indices, ) import cvxpy as cp import cvxpy.interface.matrix_utilities as intf import cvxpy.settings as s from cvxpy import Minimize, Problem from cvxpy.atoms.affine.add_expr import AddExpression from cvxpy.atoms.affine.wraps import ( hermitian_wrap, nonneg_wrap, nonpos_wrap, psd_wrap, skew_symmetric_wrap, symmetric_wrap, ) from cvxpy.expressions.constants import Constant, Parameter from cvxpy.expressions.variable import Variable from cvxpy.tests.base_test import BaseTest from cvxpy.utilities.linalg import gershgorin_psd_check class TestExpressions(BaseTest): """ Unit tests for the expression/expression module. """ def setUp(self) -> None: self.a = Variable(name='a') self.x = Variable(2, name='x') self.y = Variable(3, name='y') self.z = Variable(2, name='z') self.A = Variable((2, 2), name='A') self.B = Variable((2, 2), name='B') self.C = Variable((3, 2), name='C') self.intf = intf.DEFAULT_INTF # Test the Variable class. def test_variable(self) -> None: x = Variable(2) y = Variable(2) assert y.name() != x.name() x = Variable(2, name='x') y = Variable() self.assertEqual(x.name(), 'x') self.assertEqual(x.shape, (2,)) self.assertEqual(y.shape, tuple()) self.assertEqual(x.curvature, s.AFFINE) self.assertEqual(repr(self.x), "Variable((2,), x)") self.assertEqual(repr(self.A), "Variable((2, 2), A)") self.assertEqual(repr(cp.Variable(name='x', nonneg=True)), "Variable((), x, nonneg=True)") self.assertTrue(repr(cp.Variable()).startswith("Variable((), var")) # Test shape provided as list instead of tuple self.assertEqual(cp.Variable(shape=[2], integer=True).shape, (2,)) with self.assertRaises(Exception) as cm: Variable((2, 0)) self.assertEqual(str(cm.exception), "Invalid dimensions (2, 0).") with self.assertRaises(Exception) as cm: Variable((2, .5)) self.assertEqual(str(cm.exception), "Invalid dimensions (2, 0.5).") with self.assertRaises(Exception) as cm: Variable(2, 1) self.assertEqual(str(cm.exception), "Variable name 1 must be a string.") def test_assign_var_value(self) -> None: """Test assigning a value to a variable. """ # Scalar variable. a = Variable() a.value = 1 self.assertEqual(a.value, 1) with self.assertRaises(Exception) as cm: a.value = [2, 1] self.assertEqual(str(cm.exception), "Invalid dimensions (2,) for Variable value.") # Test assigning None. a.value = 1 a.value = None assert a.value is None # Vector variable. x = Variable(2) x.value = [2, 1] self.assertItemsAlmostEqual(x.value, [2, 1]) # Matrix variable. A = Variable((3, 2)) A.value = np.ones((3, 2)) self.assertItemsAlmostEqual(A.value, np.ones((3, 2))) # Test assigning negative val to nonnegative variable. x = Variable(nonneg=True) with self.assertRaises(Exception) as cm: x.value = -2 self.assertEqual(str(cm.exception), "Variable value must be nonnegative.") # Test tranposing variables. def test_transpose_variable(self) -> None: var = self.a.T self.assertEqual(var.name(), "a") self.assertEqual(var.shape, tuple()) self.a.save_value(2) self.assertEqual(var.value, 2) var = self.x self.assertEqual(var.name(), "x") self.assertEqual(var.shape, (2,)) x = Variable((2, 1), name='x') var = x.T self.assertEqual(var.name(), "x.T") self.assertEqual(var.shape, (1, 2)) x.save_value(np.array([[1, 2]]).T) self.assertEqual(var.value[0, 0], 1) self.assertEqual(var.value[0, 1], 2) var = self.C.T self.assertEqual(var.name(), "C.T") self.assertEqual(var.shape, (2, 3)) # coeffs = var.canonical_form[0].coefficients() # mat = coeffs.values()[0][0] # self.assertEqual(mat.shape, (2,6)) # self.assertEqual(mat[1,3], 1) index = var[1, 0] self.assertEqual(index.name(), "C.T[1, 0]") self.assertEqual(index.shape, tuple()) var = x.T.T self.assertEqual(var.name(), "x.T.T") self.assertEqual(var.shape, (2, 1)) # Test the Constant class. def test_constants(self) -> None: c = Constant(2.0) self.assertEqual(c.name(), str(2.0)) c = Constant(2) self.assertEqual(c.value, 2) self.assertEqual(c.shape, tuple()) self.assertEqual(c.curvature, s.CONSTANT) self.assertEqual(c.sign, s.NONNEG) self.assertEqual(Constant(-2).sign, s.NONPOS) self.assertEqual(Constant(0).sign, s.ZERO) # self.assertEqual(c.canonical_form[0].shape, (1, 1)) # self.assertEqual(c.canonical_form[1], []) # coeffs = c.coefficients() # self.assertEqual(coeffs.keys(), [s.CONSTANT]) # self.assertEqual(coeffs[s.CONSTANT], [2]) # Test the sign. c = Constant([[2], [2]]) self.assertEqual(c.shape, (1, 2)) self.assertEqual(c.sign, s.NONNEG) self.assertEqual((-c).sign, s.NONPOS) self.assertEqual((0*c).sign, s.ZERO) c = Constant([[2], [-2]]) self.assertEqual(c.sign, s.UNKNOWN) c = Constant(np.zeros((2, 1))) self.assertEqual(c.shape, (2, 1)) # Test sign of a complex expression. c = Constant([1, 2]) self.assertEqual(c.shape, (2,)) A = Constant([[1, 1], [1, 1]]) exp = c.T @ A @ c self.assertEqual(exp.sign, s.NONNEG) self.assertEqual((c.T @ c).sign, s.NONNEG) exp = c.T.T self.assertEqual(exp.sign, s.NONNEG) exp = c.T @ self.A self.assertEqual(exp.sign, s.UNKNOWN) # Test repr. self.assertEqual(repr(c), "Constant(CONSTANT, NONNEGATIVE, (2,))") # Test name. c = Constant(1, name="test") self.assertEqual(str(c), "test") self.assertEqual(c.name(), "test") self.assertEqual(repr(c), "Constant(CONSTANT, NONNEGATIVE, ())") def test_constant_psd_nsd(self): n = 5 np.random.randn(0) U = np.random.randn(n, n) U = U @ U.T (evals, U) = np.linalg.eigh(U) # U is now an orthogonal matrix # Try four indefinite matrices with different eigenvalue # spread around the origin. v1 = np.array([3, 2, 1, 1e-8, -1]) P = Constant(U @ np.diag(v1) @ U.T) self.assertFalse(P.is_psd()) self.assertFalse(P.is_nsd()) v2 = np.array([3, 2, 2, 1e-6, -1]) P = Constant(U @ np.diag(v2) @ U.T) self.assertFalse(P.is_psd()) self.assertFalse(P.is_nsd()) v3 = np.array([3, 2, 2, 1e-4, -1e-6]) P = Constant(U @ np.diag(v3) @ U.T) self.assertFalse(P.is_psd()) self.assertFalse(P.is_nsd()) v4 = np.array([-1, 3, 0, 0, 0]) P = Constant(U @ np.diag(v4) @ U.T) self.assertFalse(P.is_psd()) self.assertFalse(P.is_nsd()) # Try a test case given in GitHub issue 1451. # (Should be equivalent to v4 above). P = Constant(np.array([[1, 2], [2, 1]])) x = Variable(shape=(2,)) expr = cp.quad_form(x, P) self.assertFalse(expr.is_dcp()) self.assertFalse((-expr).is_dcp()) self.assertFalse(gershgorin_psd_check(P.value, tol=0.99)) # Useful Gershgorin disc check P = Constant(np.array([[2, 1], [1, 2]])) self.assertTrue(gershgorin_psd_check(P.value, tol=0.0)) # Verify good behavior for large eigenvalues P = Constant(np.diag(9*[1e-4] + [-1e4])) self.assertFalse(P.is_psd()) self.assertFalse(P.is_nsd()) # Check a case when the matrix is in fact PSD. P = Constant(np.ones(shape=(5, 5))) self.assertTrue(P.is_psd()) self.assertFalse(P.is_nsd()) # Check with sparse inputs P = Constant(sp.eye_array(10)) self.assertTrue(gershgorin_psd_check(P.value, s.EIGVAL_TOL)) self.assertTrue(P.is_psd()) self.assertTrue((-P).is_nsd()) Q = -s.EIGVAL_TOL/2 * P self.assertTrue(gershgorin_psd_check(Q.value, s.EIGVAL_TOL)) Q = -1.1*s.EIGVAL_TOL*P self.assertFalse(gershgorin_psd_check(Q.value, s.EIGVAL_TOL)) self.assertFalse(Q.is_psd()) def test_constant_skew_symmetric(self) -> None: # Define inputs M1_false = np.eye(3) M2_true = np.zeros((3, 3)) M3_true = np.array([[0, 1], [-1, 0]]) M4_true = np.array([[0, -1], [1, 0]]) M5_false = np.array([[0, 1], [1, 0]]) M6_false = np.array([[1, 1], [-1, 0]]) M7_false = np.array([[0, 1], [-1.1, 0]]) # Test dense constants C = Constant(M1_false) self.assertFalse(C.is_skew_symmetric()) C = Constant(M2_true) self.assertTrue(C.is_skew_symmetric()) C = Constant(M3_true) self.assertTrue(C.is_skew_symmetric()) C = Constant(M4_true) self.assertTrue(C.is_skew_symmetric()) C = Constant(M5_false) self.assertFalse(C.is_skew_symmetric()) C = Constant(M6_false) self.assertFalse(C.is_skew_symmetric()) C = Constant(M7_false) self.assertFalse(C.is_skew_symmetric()) # Test sparse constants C = Constant(sp.csc_array(M1_false)) self.assertFalse(C.is_skew_symmetric()) C = Constant(sp.csc_array(M2_true)) self.assertTrue(C.is_skew_symmetric()) C = Constant(sp.csc_array(M4_true)) self.assertTrue(C.is_skew_symmetric()) C = Constant(sp.csc_array(M5_false)) self.assertFalse(C.is_skew_symmetric()) C = Constant(sp.csc_array(M6_false)) self.assertFalse(C.is_skew_symmetric()) C = Constant(sp.csc_array(M7_false)) self.assertFalse(C.is_skew_symmetric()) # Test complex inputs: never recognized as skew-symmetric. C = Constant(1j * M2_true) self.assertFalse(C.is_skew_symmetric()) # ^ From a mathematical standpoint one can argue that this should # be true, but I don't think there's precedent for CVXPY # automatically converting complex expressions with zero imaginary-part # into equivalent real expressions. -- Riley C = Constant(1j * M3_true) self.assertFalse(C.is_skew_symmetric()) C = Constant(1j * M4_true) self.assertFalse(C.is_skew_symmetric()) pass def test_1D_array(self) -> None: """Test NumPy 1D arrays as constants. """ c = np.array([1, 2]) p = Parameter(2) p.value = [1, 1] self.assertEqual((c @ p).value, 3) self.assertEqual((c @ self.x).shape, tuple()) # Test Parameter class on good inputs. def test_parameters_successes(self) -> None: # Parameter names and dimensions p = Parameter(name='p') self.assertEqual(p.name(), "p") self.assertEqual(p.shape, tuple()) # Entry-wise constraints on parameter values. val = -np.ones((4, 3)) val[0, 0] = 2 p = Parameter((4, 3)) p.value = val # Initialize a parameter with a value; later, set it to None. p = Parameter(value=10) self.assertEqual(p.value, 10) p.value = 10 p.value = None self.assertEqual(p.value, None) # Test parameter representation. p = Parameter((4, 3), nonpos=True) self.assertEqual(repr(p), 'Parameter((4, 3), nonpos=True)') # Test valid diagonal parameter. p = Parameter((2, 2), diag=True) p.value = sp.csc_array(np.eye(2)) self.assertItemsAlmostEqual(p.value.todense(), np.eye(2), places=10) def test_psd_nsd_parameters(self) -> None: # Test valid rank-deficeint PSD parameter. np.random.seed(42) a = np.random.normal(size=(100, 95)) a2 = a.dot(a.T) # This must be a PSD matrix. p = Parameter((100, 100), PSD=True) p.value = a2 self.assertItemsAlmostEqual(p.value, a2, places=10) # Test positive definite matrix with non-distinct eigenvalues m, n = 10, 5 A = np.random.randn(m, n) + 1j * np.random.randn(m, n) # a random complex matrix A = np.dot(A.T.conj(), A) # a random Hermitian positive definite matrix A = np.vstack([np.hstack([np.real(A), -np.imag(A)]), np.hstack([np.imag(A), np.real(A)])]) p = Parameter(shape=(2*n, 2*n), PSD=True) p.value = A self.assertItemsAlmostEqual(p.value, A) # Test arithmetic. p = Parameter(shape=(2, 2), PSD=True) self.assertTrue((2*p).is_psd()) self.assertTrue((p + p).is_psd()) self.assertTrue((-p).is_nsd()) self.assertTrue(((-2)*(-p)).is_psd()) # Test invalid PSD and NSD parameters n = 5 P = Parameter(shape=(n, n), PSD=True) N = Parameter(shape=(n, n), NSD=True) np.random.randn(0) U = np.random.randn(n, n) U = U @ U.T (evals, U) = np.linalg.eigh(U) # U is now an orthogonal matrix v1 = np.array([3, 2, 1, 1e-8, -1]) v2 = np.array([3, 2, 2, 1e-6, -1]) v3 = np.array([3, 2, 2, 1e-4, -1e-6]) v4 = np.array([-1, 3, 0, 0, 0]) vs = [v1, v2, v3, v4] for vi in vs: with self.assertRaises(Exception) as cm: P.value = U @ np.diag(vi) @ U.T self.assertEqual(str(cm.exception), "Parameter value must be positive semidefinite.") with self.assertRaises(Exception) as cm: N.value = -U @ np.diag(vi) @ U.T self.assertEqual(str(cm.exception), "Parameter value must be negative semidefinite.") # Test the Parameter class on bad inputs. def test_parameters_failures(self) -> None: p = Parameter(name='p') self.assertEqual(p.name(), "p") self.assertEqual(p.shape, tuple()) p = Parameter((4, 3), nonneg=True) with self.assertRaises(Exception) as cm: p.value = 1 self.assertEqual(str(cm.exception), "Invalid dimensions () for Parameter value.") val = -np.ones((4, 3)) val[0, 0] = 2 p = Parameter((4, 3), nonneg=True) with self.assertRaises(Exception) as cm: p.value = val self.assertEqual(str(cm.exception), "Parameter value must be nonnegative.") p = Parameter((4, 3), nonpos=True) with self.assertRaises(Exception) as cm: p.value = val self.assertEqual(str(cm.exception), "Parameter value must be nonpositive.") with self.assertRaises(Exception) as cm: p = Parameter(2, 1, nonpos=True, value=[2, 1]) self.assertEqual(str(cm.exception), "Parameter value must be nonpositive.") with self.assertRaises(Exception) as cm: p = Parameter((4, 3), nonneg=True, value=[1, 2]) self.assertEqual(str(cm.exception), "Invalid dimensions (2,) for Parameter value.") # Boolean with self.assertRaises(Exception) as cm: p = Parameter((2, 2), boolean=True, value=[[1, 1], [1, -1]]) self.assertEqual(str(cm.exception), "Parameter value must be boolean.") # Integer with self.assertRaises(Exception) as cm: p = Parameter((2, 2), integer=True, value=[[1, 1.5], [1, -1]]) self.assertEqual(str(cm.exception), "Parameter value must be integer.") # Boolean indices with self.assertRaises(Exception) as cm: p = Parameter((2, 2), boolean=[(0, 0), (0, 1)], value=[[0, 2], [1, 0]]) self.assertEqual(str(cm.exception), "Parameter value must be boolean.") # Integer indices with self.assertRaises(Exception) as cm: p = Parameter((2, 2), integer=[(0, 0), (0, 1)], value=[[1, 1.5], [1.4, 2.8]]) self.assertEqual(str(cm.exception), "Parameter value must be integer.") # Diag. with self.assertRaises(Exception) as cm: p = Parameter((2, 2), diag=True, value=[[1, 1], [1, -1]]) self.assertEqual(str(cm.exception), "Parameter value must be diagonal.") # Symmetric. with self.assertRaises(Exception) as cm: p = Parameter((2, 2), symmetric=True, value=[[1, 1], [-1, -1]]) self.assertEqual(str(cm.exception), "Parameter value must be symmetric.") def test_symmetric(self) -> None: """Test symmetric variables. """ with self.assertRaises(Exception) as cm: v = Variable((4, 3), symmetric=True) self.assertEqual(str(cm.exception), "Invalid dimensions (4, 3). Must be a square matrix.") v = Variable((2, 2), symmetric=True) assert v.is_symmetric() v = Variable((2, 2), PSD=True) assert v.is_symmetric() v = Variable((2, 2), NSD=True) assert v.is_symmetric() v = Variable((2, 2), diag=True) assert v.is_symmetric() assert self.a.is_symmetric() assert not self.A.is_symmetric() v = Variable((2, 2), symmetric=True) expr = v + v assert expr.is_symmetric() expr = -v assert expr.is_symmetric() expr = v.T assert expr.is_symmetric() expr = cp.real(v) assert expr.is_symmetric() expr = cp.imag(v) assert expr.is_symmetric() expr = cp.conj(v) assert expr.is_symmetric() expr = cp.promote(Variable(), (2, 2)) assert expr.is_symmetric() def test_hermitian(self) -> None: """Test Hermitian variables. """ with self.assertRaises(Exception) as cm: v = Variable((4, 3), hermitian=True) self.assertEqual(str(cm.exception), "Invalid dimensions (4, 3). Must be a square matrix.") v = Variable((2, 2), hermitian=True) assert v.is_hermitian() # v = Variable((2,2), PSD=True) # assert v.is_symmetric() # v = Variable((2,2), NSD=True) # assert v.is_symmetric() v = Variable((2, 2), diag=True) assert v.is_hermitian() v = Variable((2, 2), hermitian=True) expr = v + v assert expr.is_hermitian() expr = -v assert expr.is_hermitian() expr = v.T assert expr.is_hermitian() expr = cp.real(v) assert expr.is_hermitian() expr = cp.imag(v) assert expr.is_hermitian() expr = cp.conj(v) assert expr.is_hermitian() expr = cp.promote(Variable(), (2, 2)) assert expr.is_hermitian() def test_round_attr(self) -> None: """Test rounding for attributes. """ # Nonpos v = Variable(1, nonpos=True) self.assertAlmostEqual(v.project(1), 0) v = Variable(2, nonpos=True) self.assertItemsAlmostEqual(v.project(np.array([1, -1])), [0, -1]) # Nonneg v = Variable(1, nonneg=True) self.assertAlmostEqual(v.project(-1), 0) v = Variable(2, nonneg=True) self.assertItemsAlmostEqual(v.project(np.array([1, -1])), [1, 0]) # Boolean v = Variable((2, 2), boolean=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1], [1, 0]]).T), [1, 0, 1, 0]) # Integer v = Variable((2, 2), integer=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1.6], [1, 0]]).T), [1, -2, 1, 0]) # Symmetric v = Variable((2, 2), symmetric=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1], [1, 0]])), [1, 0, 0, 0]) # PSD v = Variable((2, 2), PSD=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1], [1, -1]])), [1, 0, 0, 0]) # NSD v = Variable((2, 2), NSD=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1], [1, -1]])), [0, 0, 0, -1]) # diag v = Variable((2, 2), diag=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1], [1, 0]])).todense(), [1, 0, 0, 0]) # Hermitian v = Variable((2, 2), hermitian=True) self.assertItemsAlmostEqual(v.project(np.array([[1, -1j], [1, 0]])), [1, 0.5+0.5j, 0.5-0.5j, 0]) A = Constant(np.array([[1.0]])) self.assertEqual(A.is_psd(), True) self.assertEqual(A.is_nsd(), False) A = Constant(np.array([[-1.0]])) self.assertEqual(A.is_psd(), False) self.assertEqual(A.is_nsd(), True) A = Constant(np.array([[0.0]])) self.assertEqual(A.is_psd(), True) self.assertEqual(A.is_nsd(), True) def test_project_boolean_indices(self) -> None: idx = (np.array([0, 2]),) leaf = cp.Variable((3,), boolean=idx) val = np.array([-0.2, 0.5, 1.7]) projected = leaf.project(val) # Only indices 0 and 2 are projected to boolean, index 1 is unchanged assert (projected == np.array([0, 0.5, 1])).all() def test_project_integer_indices(self) -> None: idx = (np.array([1]),) leaf = cp.Variable((3,), integer=idx) val = np.array([1.2, 2.7, -0.8]) projected = leaf.project(val) # Only index 1 is projected to integer, others unchanged assert (projected == np.array([1.2, 3, -0.8])).all() def test_add_expression(self) -> None: # Vectors c = Constant([2, 2]) exp = self.x + c self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.sign, s.UNKNOWN) # self.assertEqual(exp.canonical_form[0].shape, (2, 1)) # self.assertEqual(exp.canonical_form[1], []) # self.assertEqual(exp.name(), self.x.name() + " + " + c.name()) self.assertEqual(exp.shape, (2,)) z = Variable(2, name='z') exp = exp + z + self.x # Incompatible dimensions with self.assertRaises(ValueError): (self.x + self.y) # Matrices exp = self.A + self.B self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (2, 2)) # Incompatible dimensions with self.assertRaises(ValueError): (self.A + self.C) # Incompatible dimensions with self.assertRaises(ValueError): AddExpression([self.A, self.C]) # Test that sum is flattened. exp = self.x + c + self.x self.assertEqual(len(exp.args), 3) # Test repr. self.assertEqual(repr(exp), "Expression(AFFINE, UNKNOWN, (2,))") # Test the SubExpresion class. def test_sub_expression(self) -> None: # Vectors c = Constant([2, 2]) exp = self.x - c self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.sign, s.UNKNOWN) # self.assertEqual(exp.canonical_form[0].shape, (2, 1)) # self.assertEqual(exp.canonical_form[1], []) # self.assertEqual(exp.name(), self.x.name() + " - " + Constant([2,2]).name()) self.assertEqual(exp.shape, (2,)) z = Variable(2, name='z') exp = exp - z - self.x # Incompatible dimensions with self.assertRaises(ValueError): (self.x - self.y) # Matrices exp = self.A - self.B self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (2, 2)) # Incompatible dimensions with self.assertRaises(ValueError): (self.A - self.C) # Test repr. self.assertEqual(repr(self.x - c), "Expression(AFFINE, UNKNOWN, (2,))") # Test the MulExpresion class. def test_mul_expression(self) -> None: # Vectors c = Constant([[2], [2]]) exp = c @ self.x self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual((c[0] @ self.x).sign, s.UNKNOWN) # self.assertEqual(exp.canonical_form[0].shape, (1, 1)) # self.assertEqual(exp.canonical_form[1], []) # self.assertEqual(exp.name(), c.name() + " * " + self.x.name()) self.assertEqual(exp.shape, (1,)) # Incompatible dimensions with self.assertRaises(ValueError): ([2, 2, 3] @ self.x) # Matrices: Incompatible dimensions with self.assertRaises(ValueError): Constant([[2, 1], [2, 2]]) @ self.C # Affine times affine is okay with warnings.catch_warnings(): warnings.simplefilter("ignore") q = self.A @ self.B self.assertTrue(q.is_quadratic()) # Constant expressions T = Constant([[1, 2, 3], [3, 5, 5]]) exp = (T + T) @ self.B self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (3, 2)) # Expression that would break sign multiplication without promotion. c = Constant([[2], [2], [-2]]) exp = [[1], [2]] + c @ self.C self.assertEqual(exp.sign, s.UNKNOWN) # By default, warnings are raised if we access matmul from *. c = Constant([[2], [2]]) with warnings.catch_warnings(record=True) as w: c * self.x self.assertEqual(2, len(w)) self.assertEqual(w[0].category, UserWarning) self.assertEqual(w[1].category, DeprecationWarning) # repeat, to make sure warnings continue to be displayed c * self.x self.assertEqual(4, len(w)) self.assertEqual(w[2].category, UserWarning) self.assertEqual(w[3].category, DeprecationWarning) # suppress one of the two warnings warnings.simplefilter('ignore', DeprecationWarning) c * self.x self.assertEqual(5, len(w)) # suppress both warnings warnings.simplefilter('ignore', UserWarning) c * self.x self.assertEqual(len(w), 5) # verify that an error can be raised. warnings.simplefilter("error", UserWarning) with self.assertRaises(UserWarning): c * self.x def test_matmul_expression(self) -> None: """Test matmul function, corresponding to .__matmul__( operator. """ # Vectors c = Constant([[2], [2]]) exp = c.__matmul__(self.x) self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.sign, s.UNKNOWN) # self.assertEqual(exp.name(), c.name() + " .__matmul__( " + self.x.name()) self.assertEqual(exp.shape, (1,)) with self.assertRaises(Exception) as cm: self.x.__matmul__(2) self.assertEqual(str(cm.exception), "Scalar operands are not allowed, use '*' instead") # Incompatible dimensions with self.assertRaises(ValueError) as cm: (self.x.__matmul__(np.array([2, 2, 3]))) # Incompatible dimensions with self.assertRaises(Exception) as cm: Constant([[2, 1], [2, 2]]) .__matmul__(self.C) # Affine times affine is okay with warnings.catch_warnings(): warnings.simplefilter("ignore") q = self.A .__matmul__(self.B) self.assertTrue(q.is_quadratic()) # Constant expressions T = Constant([[1, 2, 3], [3, 5, 5]]) exp = (T + T) .__matmul__(self.B) self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (3, 2)) # Expression that would break sign multiplication without promotion. c = Constant([[2], [2], [-2]]) exp = [[1], [2]] + c.__matmul__(self.C) self.assertEqual(exp.sign, s.UNKNOWN) # Testing shape. a = Parameter((1,)) x = Variable(shape=(1,)) expr = a.__matmul__(x) self.assertEqual(expr.shape, ()) # Testing shape. a = Parameter((1,)) x = Variable(shape=(1,)) expr = a.__matmul__(x) self.assertEqual(expr.shape, ()) A = Parameter((4, 4)) z = Variable((4, 1)) expr = A.__matmul__(z) self.assertEqual(expr.shape, (4, 1)) v = Variable((1, 1)) col_scalar = Parameter((1, 1)) assert v.shape == col_scalar.shape == col_scalar.T.shape # Test the DivExpresion class. def test_div_expression(self) -> None: # Vectors exp = self.x/2 self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.sign, s.UNKNOWN) # self.assertEqual(exp.canonical_form[0].shape, (2, 1)) # self.assertEqual(exp.canonical_form[1], []) # self.assertEqual(exp.name(), c.name() + " * " + self.x.name()) self.assertEqual(exp.shape, (2,)) with self.assertRaises(Exception) as cm: (self.x/[2, 2, 3]) print(cm.exception) self.assertRegex(str(cm.exception), "Incompatible shapes for division") c = Constant([3.0, 4.0, 12.0]) self.assertItemsAlmostEqual( (c / Constant([1.0, 2.0, 3.0])).value, np.array([3.0, 2.0, 4.0])) # Constant expressions. c = Constant(2) exp = c/(3 - 5) self.assertEqual(exp.curvature, s.CONSTANT) self.assertEqual(exp.shape, tuple()) self.assertEqual(exp.sign, s.NONPOS) # Parameters. p = Parameter(nonneg=True) exp = 2/p p.value = 2 self.assertEqual(exp.value, 1) rho = Parameter(nonneg=True) rho.value = 1 self.assertEqual(rho.sign, s.NONNEG) self.assertEqual(Constant(2).sign, s.NONNEG) self.assertEqual((Constant(2)/Constant(2)).sign, s.NONNEG) self.assertEqual((Constant(2)*rho).sign, s.NONNEG) self.assertEqual((rho/2).sign, s.NONNEG) # Broadcasting. x = cp.Variable((3, 3)) c = np.arange(1, 4)[:, None] expr = x / c self.assertEqual((3, 3), expr.shape) x.value = np.ones((3, 3)) A = np.ones((3, 3)) / c self.assertItemsAlmostEqual(A, expr.value) # Test the NegExpression class. def test_neg_expression(self) -> None: # Vectors exp = -self.x self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (2,)) assert exp.is_affine() self.assertEqual(exp.sign, s.UNKNOWN) assert not exp.is_nonneg() self.assertEqual(exp.shape, self.x.shape) self.assertEqual(exp.name(), "-%s" % self.x.name()) # Matrices exp = -self.C self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (3, 2)) # String b = cp.Variable(name='b') X = cp.Variable((2,1), name='X') A = cp.Variable((1,2), name='A') exp = -(A@X - b) self.assertEqual(str(exp), "-(A @ X + -b)") # Test promotion of scalar constants. def test_scalar_const_promotion(self) -> None: # Vectors exp = self.x + 2 self.assertEqual(exp.curvature, s.AFFINE) assert exp.is_affine() self.assertEqual(exp.sign, s.UNKNOWN) assert not exp.is_nonpos() # self.assertEqual(exp.canonical_form[0].shape, (2, 1)) # self.assertEqual(exp.canonical_form[1], []) # self.assertEqual(exp.name(), self.x.name() + " + " + Constant(2).name()) self.assertEqual(exp.shape, (2,)) self.assertEqual((4 - self.x).shape, (2,)) self.assertEqual((4 * self.x).shape, (2,)) self.assertEqual((4 <= self.x).shape, (2,)) self.assertEqual((4 == self.x).shape, (2,)) self.assertEqual((self.x >= 4).shape, (2,)) # Matrices exp = (self.A + 2) + 4 self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual((3 * self.A).shape, (2, 2)) self.assertEqual(exp.shape, (2, 2)) # Test indexing expression. def test_index_expression(self) -> None: # Tuple of integers as key. exp = self.x[1] # self.assertEqual(exp.name(), "x[1,0]") self.assertEqual(exp.curvature, s.AFFINE) assert exp.is_affine() self.assertEqual(exp.shape, tuple()) # coeff = exp.canonical_form[0].coefficients()[self.x][0] # self.assertEqual(coeff[0,1], 1) self.assertEqual(exp.value, None) exp = self.x[1].T # self.assertEqual(exp.name(), "x[1,0]") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, tuple()) with self.assertRaises(Exception) as cm: (self.x[2, 0]) self.assertEqual(str(cm.exception), "Too many indices for expression.") with self.assertRaises(Exception) as cm: (self.x[2]) self.assertEqual(str(cm.exception), "Index 2 is out of bounds for axis 0 with size 2.") # Slicing exp = self.C[0:2, 1] # self.assertEqual(exp.name(), "C[0:2,1]") self.assertEqual(exp.shape, (2,)) exp = self.C[0:, 0:2] # self.assertEqual(exp.name(), "C[0:,0:2]") self.assertEqual(exp.shape, (3, 2)) exp = self.C[0::2, 0::2] # self.assertEqual(exp.name(), "C[0::2,0::2]") self.assertEqual(exp.shape, (2, 1)) exp = self.C[:3, :1:2] # self.assertEqual(exp.name(), "C[0:3,0]") self.assertEqual(exp.shape, (3, 1)) exp = self.C[0:, 0] # self.assertEqual(exp.name(), "C[0:,0]") self.assertEqual(exp.shape, (3,)) c = Constant([[1, -2], [0, 4]]) exp = c[1, 1] self.assertEqual(exp.curvature, s.CONSTANT) self.assertEqual(exp.sign, s.UNKNOWN) self.assertEqual(c[0, 1].sign, s.UNKNOWN) self.assertEqual(c[1, 0].sign, s.UNKNOWN) self.assertEqual(exp.shape, tuple()) self.assertEqual(exp.value, 4) c = Constant([[1, -2, 3], [0, 4, 5], [7, 8, 9]]) exp = c[0:3, 0:4:2] self.assertEqual(exp.curvature, s.CONSTANT) assert exp.is_constant() self.assertEqual(exp.shape, (3, 2)) self.assertEqual(exp[0, 1].value, 7) # Slice of transpose exp = self.C.T[0:2, 1:2] self.assertEqual(exp.shape, (2, 1)) # Arithmetic expression indexing exp = (self.x + self.z)[1] # self.assertEqual(exp.name(), "x[1,0] + z[1,0]") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.sign, s.UNKNOWN) self.assertEqual(exp.shape, tuple()) exp = (self.x + self.a)[1:2] # self.assertEqual(exp.name(), "x[1,0] + a") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (1,)) exp = (self.x - self.z)[1:2] # self.assertEqual(exp.name(), "x[1,0] - z[1,0]") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (1,)) exp = (self.x - self.a)[1] # self.assertEqual(exp.name(), "x[1,0] - a") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, tuple()) exp = (-self.x)[1] # self.assertEqual(exp.name(), "-x[1,0]") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, tuple()) c = Constant([[1, 2], [3, 4]]) exp = (c @ self.x)[1] # self.assertEqual(exp.name(), "[[2], [4]] * x[0:,0]") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, tuple()) c = Constant([[1, 2], [3, 4]]) exp = (c * self.a)[1, 0:1] # self.assertEqual(exp.name(), "2 * a") self.assertEqual(exp.curvature, s.AFFINE) self.assertEqual(exp.shape, (1,)) def test_special_idx_str_repr(self) -> None: idx = [i for i in range(178)] exp = cp.Variable((200, 10), name="exp")[idx, 6] self.assertEqual("exp[[0, 1, 2, ..., 175, 176, 177], 6]", str(exp)) idx = [i for i in range(5)] exp = cp.Variable((10, 10), name="exp")[idx, 2:5] self.assertEqual("exp[[0, 1, 2, 3, 4], 2:5]", str(exp)) def test_none_idx(self) -> None: """Test None as index. """ expr = self.a[None, None] self.assertEqual(expr.shape, (1, 1)) expr = self.x[:, None] self.assertEqual(expr.shape, (2, 1)) expr = self.x[None, :] self.assertEqual(expr.shape, (1, 2)) expr = Constant([1, 2])[None, :] self.assertEqual(expr.shape, (1, 2)) self.assertItemsAlmostEqual(expr.value, [1, 2]) def test_out_of_bounds(self) -> None: """Test out of bounds indices. """ with self.assertRaises(Exception) as cm: self.x[100] self.assertEqual(str(cm.exception), "Index 100 is out of bounds for axis 0 with size 2.") with self.assertRaises(Exception) as cm: self.x[-100] self.assertEqual(str(cm.exception), "Index -100 is out of bounds for axis 0 with size 2.") exp = self.x[:-100] self.assertEqual(exp.size, 0) self.assertItemsAlmostEqual(exp.value, np.array([])) exp = self.C[100:2] self.assertEqual(exp.shape, (0, 2)) exp = self.C[:, -199:2] self.assertEqual(exp.shape, (3, 2)) exp = self.C[:, -199:-3] self.assertEqual(exp.shape, (3, 0)) def test_float_is_invalid_index(self) -> None: with self.assertRaises(IndexError) as cm: self.x[1.0] self.assertEqual(str(cm.exception), "float is an invalid index type.") with self.assertRaises(IndexError) as cm: self.x[(1.0,)] self.assertEqual(str(cm.exception), "float is an invalid index type.") with self.assertRaises(IndexError) as cm: self.C[: 2.:40] self.assertEqual(str(cm.exception), "float is an invalid index type.") with self.assertRaises(IndexError) as cm: self.x[np.array([1.0, 2.0])] self.assertEqual(str(cm.exception), "arrays used as indices must be of integer (or boolean) type") def test_neg_indices(self) -> None: """Test negative indices. """ c = Constant([[1, 2], [3, 4]]) exp = c[-1, -1] self.assertEqual(exp.value, 4) self.assertEqual(exp.shape, tuple()) self.assertEqual(exp.curvature, s.CONSTANT) c = Constant([1, 2, 3, 4]) exp = c[1:-1] self.assertItemsAlmostEqual(exp.value, [2, 3]) self.assertEqual(exp.shape, (2,)) self.assertEqual(exp.curvature, s.CONSTANT) c = Constant([1, 2, 3, 4]) exp = c[::-1] self.assertItemsAlmostEqual(exp.value, [4, 3, 2, 1]) self.assertEqual(exp.shape, (4,)) self.assertEqual(exp.curvature, s.CONSTANT) x = Variable(4) self.assertEqual(x[::-1].shape, (4,)) Problem(Minimize(0), [x[::-1] == c]).solve(solver=cp.SCS) self.assertItemsAlmostEqual(x.value, [4, 3, 2, 1]) x = Variable(2) self.assertEqual(x[::-1].shape, (2,)) x = Variable(100, name="x") self.assertEqual("x[0:99]", str(x[:-1])) c = Constant([[1, 2], [3, 4]]) expr = c[0, 2:0:-1] self.assertEqual(expr.shape, (1,)) self.assertAlmostEqual(expr.value, 3) expr = c[0, 2::-1] self.assertEqual(expr.shape, (2,)) self.assertItemsAlmostEqual(expr.value, [3, 1]) def test_logical_indices(self) -> None: """Test indexing with boolean arrays. """ A = np.array([[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12]]) C = Constant(A) # Boolean array. expr = C[A <= 2] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[A <= 2], expr.value) expr = C[A % 2 == 0] self.assertEqual(expr.shape, (6,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[A % 2 == 0], expr.value) # Boolean array for rows, index for columns. expr = C[np.array([True, False, True]), 3] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[np.array([True, False, True]), 3], expr.value) # Index for row, boolean array for columns. expr = C[1, np.array([True, False, False, True])] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[1, np.array([True, False, False, True])], expr.value) # Boolean array for rows, slice for columns. expr = C[np.array([True, True, True]), 1:3] self.assertEqual(expr.shape, (3, 2)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[np.array([True, True, True]), 1:3], expr.value) # Slice for row, boolean array for columns. expr = C[1:-1, np.array([True, False, True, True])] self.assertEqual(expr.shape, (1, 3)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[1:-1, np.array([True, False, True, True])], expr.value) # Boolean arrays for rows and columns. # Not sure what this does. expr = C[np.array([True, True, True]), np.array([True, False, True, True])] self.assertEqual(expr.shape, (3,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[np.array([True, True, True]), np.array([True, False, True, True])], expr.value) def test_selector_list_indices(self) -> None: """Test indexing with lists/ndarrays of indices. """ A = np.array([[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12]]) C = Constant(A) # List for rows. expr = C[[1, 2]] self.assertEqual(expr.shape, (2, 4)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[[1, 2]], expr.value) # List for rows, index for columns. expr = C[[0, 2], 3] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[[0, 2], 3], expr.value) # Index for row, list for columns. expr = C[1, [0, 2]] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[1, [0, 2]], expr.value) # List for rows, slice for columns. expr = C[[0, 2], 1:3] self.assertEqual(expr.shape, (2, 2)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[[0, 2], 1:3], expr.value) # Slice for row, list for columns. expr = C[1:-1, [0, 2]] self.assertEqual(expr.shape, (1, 2)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[1:-1, [0, 2]], expr.value) # Lists for rows and columns. expr = C[[0, 1], [1, 3]] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[[0, 1], [1, 3]], expr.value) # Ndarray for rows, list for columns. expr = C[np.array([0, 1]), [1, 3]] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[np.array([0, 1]), [1, 3]], expr.value) # Ndarrays for rows and columns. expr = C[np.array([0, 1]), np.array([1, 3])] self.assertEqual(expr.shape, (2,)) self.assertEqual(expr.sign, s.NONNEG) self.assertItemsAlmostEqual(A[np.array([0, 1]), np.array([1, 3])], expr.value) def test_powers(self) -> None: exp = self.x**2 self.assertEqual(exp.curvature, s.CONVEX) exp = self.x**0.5 self.assertEqual(exp.curvature, s.CONCAVE) exp = self.x**-1 self.assertEqual(exp.curvature, s.CONVEX) def test_sum(self) -> None: """Test cvxpy sum function. """ self.a.value = 1 expr = cp.sum(self.a) self.assertEqual(expr.value, 1) self.x.value = [1, 2] expr = cp.sum(self.x) self.assertEqual(expr.value, 3) def test_var_copy(self) -> None: """Test the copy function for variable types. """ x = Variable((3, 4), name="x") y = x.copy() self.assertEqual(y.shape, (3, 4)) self.assertEqual(y.name(), "x") x = Variable((5, 5), PSD=True, name="x") y = x.copy() self.assertEqual(y.shape, (5, 5)) def test_param_copy(self) -> None: """Test the copy function for Parameters. """ x = Parameter((3, 4), name="x", nonneg=True) y = x.copy() self.assertEqual(y.shape, (3, 4)) self.assertEqual(y.name(), "x") self.assertEqual(y.sign, "NONNEGATIVE") def test_constant_copy(self) -> None: """Test the copy function for Constants. """ x = Constant(2) y = x.copy() self.assertEqual(y.shape, tuple()) self.assertEqual(y.value, 2) def test_is_pwl(self) -> None: """Test is_pwl() """ A = np.ones((2, 3)) b = np.ones(2) expr = A @ self.y - b self.assertEqual(expr.is_pwl(), True) expr = cp.maximum(1, 3 * self.y) self.assertEqual(expr.is_pwl(), True) expr = cp.abs(self.y) self.assertEqual(expr.is_pwl(), True) expr = cp.pnorm(3 * self.y, 1) self.assertEqual(expr.is_pwl(), True) expr = cp.pnorm(3 * self.y ** 2, 1) self.assertEqual(expr.is_pwl(), False) def test_broadcast_mul(self) -> None: """Test multiply broadcasting. """ y = Parameter((3, 1)) z = Variable((1, 3)) y.value = np.arange(3)[:, None] z.value = (np.arange(3) - 1)[None, :] expr = cp.multiply(y, z) self.assertItemsAlmostEqual(expr.value, y.value * z.value) prob = cp.Problem(cp.Minimize(cp.sum(expr)), [z == z.value]) prob.solve(solver=cp.SCS) self.assertItemsAlmostEqual(expr.value, y.value * z.value) np.random.seed(0) m, n = 3, 4 A = np.random.rand(m, n) col_scale = Variable(n) col_scale = Variable([1, n]) C = cp.multiply(A, col_scale) self.assertEqual(C.shape, (m, n)) row_scale = Variable([m, 1]) R = cp.multiply(A, row_scale) self.assertEqual(R.shape, (m, n)) def test_broadcast_add(self) -> None: """Test addition broadcasting. """ y = Parameter((3, 1)) z = Variable((1, 3)) y.value = np.arange(3)[:, None] z.value = (np.arange(3) - 1)[None, :] expr = y + z self.assertItemsAlmostEqual(expr.value, y.value + z.value) prob = cp.Problem(cp.Minimize(cp.sum(expr)), [z == z.value]) prob.solve(solver=cp.SCS) self.assertItemsAlmostEqual(expr.value, y.value + z.value) np.random.seed(0) m, n = 3, 4 A = np.random.rand(m, n) col_scale = Variable(n) col_scale = Variable([1, n]) C = A + col_scale self.assertEqual(C.shape, (m, n)) row_scale = Variable([m, 1]) R = A + row_scale self.assertEqual(R.shape, (m, n)) def test_curvatures(self) -> None: """Test the curvatures property returns a list of valid curvatures""" mat = np.array([[1, -1]]) self.assertEqual(cp.sum(mat @ cp.square(Variable(2))).curvatures, [s.UNKNOWN]) y = cp.Constant(42) self.assertIn(s.CONSTANT, y.curvatures) x = cp.Variable(pos=True) self.assertEqual(x.curvatures, [s.AFFINE, s.CONVEX, s.CONCAVE, s.LOG_LOG_AFFINE, s.LOG_LOG_CONVEX, s.LOG_LOG_CONCAVE, s.QUASILINEAR, s.QUASICONVEX, s.QUASICONCAVE]) monomial = x*x*x self.assertEqual(monomial.curvatures, [s.LOG_LOG_AFFINE, s.LOG_LOG_CONVEX, s.LOG_LOG_CONCAVE]) posynomial = x*x*x + x self.assertEqual(posynomial.curvatures, [s.LOG_LOG_CONVEX]) llcv = 1/(x*x*x + x) self.assertEqual(llcv.curvatures, [s.LOG_LOG_CONCAVE]) def test_log_log_curvature(self) -> None: """Test that the curvature string is populated for log-log expressions. """ x = Variable(pos=True) monomial = x*x*x assert monomial.curvature == s.LOG_LOG_AFFINE posynomial = x*x*x + x assert posynomial.curvature == s.LOG_LOG_CONVEX llcv = 1/(x*x*x + x) assert llcv.curvature == s.LOG_LOG_CONCAVE def test_quad_form_matmul(self) -> None: """Test conversion of native x.T @ A @ x into QuadForms. """ # Trivial quad form x = Variable(shape=(2,)) A = Constant([[1, 0], [0, -1]]) expr = x.T.__matmul__(A).__matmul__(x) assert isinstance(expr, cp.QuadForm) # QuadForm inside nested expr: 0.5 * (x.T @ A @ x) + x.T @ x x = Variable(shape=(2,)) A = Constant([[1, 0], [0, -1]]) expr = (1 / 2) * (x.T.__matmul__(A).__matmul__(x)) + x.T.__matmul__(x) assert isinstance(expr.args[0].args[1], cp.QuadForm) assert expr.args[0].args[1].args[0] is x # QuadForm inside nested expr: (0.5 * c.T @ c) * (x.T @ A @ x) + x.T @ x x = Variable(shape=(2,)) A = Constant([[1, 0], [0, -1]]) c = Constant([2, -2]) expr = (1 / 2 * c.T.__matmul__(c)) * (x.T.__matmul__(A).__matmul__(x)) + x.T.__matmul__(x) assert isinstance(expr.args[0].args[1], cp.QuadForm) assert expr.args[0].args[1].args[0] is x # QuadForm with sparse matrices x = Variable(shape=(2,)) A = Constant(sp.eye_array(2)) expr = x.T.__matmul__(A).__matmul__(x) assert isinstance(expr, cp.QuadForm) # QuadForm with mismatched dimensions raises error x = Variable(shape=(2,)) A = Constant(np.eye(3)) with self.assertRaises(Exception) as _: x.T.__matmul__(A).__matmul__(x) # QuadForm with PSD-wrapped matrix x = cp.Variable(shape=(2,)) A = cp.Constant([[1, 0], [0, 1]]) expr = x.T.__matmul__(psd_wrap(A)).__matmul__(x) assert isinstance(expr, cp.QuadForm) # QuadForm with nested subexpr x = cp.Variable(shape=(2,)) A = cp.Constant([[2, 0, 0], [0, 0, 1]]) M = cp.Constant([[2, 0, 0], [0, 2, 0], [0, 0, 2]]) b = cp.Constant([1, 2, 3]) y = A.__matmul__(x) - b expr = y.T.__matmul__(M).__matmul__(y) assert isinstance(expr, cp.QuadForm) assert expr.args[0] is y assert expr.args[1] is M # QuadForm with parameters x = Variable(shape=(2,)) A = Parameter(shape=(2, 2), symmetric=True) expr = x.T.__matmul__(A).__matmul__(x) assert isinstance(expr, cp.QuadForm) # Expect error for asymmetric/nonhermitian matrices x = Variable(shape=(2,)) A = Constant([[1, 0], [1, 1]]) with self.assertRaises(ValueError) as _: x.T.__matmul__(A).__matmul__(x) x = Variable(shape=(2,)) A = Constant([[1, 1j], [1j, 1]]) with self.assertRaises(ValueError) as _: x.T.__matmul__(A).__matmul__(x) # Not a quad_form because x.T @ A @ y where x, y not necessarily equal x = Variable(shape=(2,)) y = Variable(shape=(2,)) A = Constant([[1, 0], [0, -1]]) expr = x.T.__matmul__(A).__matmul__(y) assert not isinstance(expr, cp.QuadForm) # Not a quad_form because M is variable x = Variable(shape=(2,)) M = Variable(shape=(2, 2)) expr = x.T.__matmul__(M).__matmul__(x) assert not isinstance(expr, cp.QuadForm) x = Constant([1, 0]) M = Variable(shape=(2, 2)) expr = x.T.__matmul__(M).__matmul__(x) assert not isinstance(expr, cp.QuadForm) def test_matmul_scalars(self) -> None: """Test evaluating a matmul that reduces one argument internally to a scalar. """ x = cp.Variable((2,)) quad = cp.quad_form(x, np.eye(2)) a = np.array([2]) expr = quad * a x.value = np.array([1, 2]) P = np.eye(2) true_val = (np.transpose(x.value) @ P @ x.value) * a assert quad.shape == () self.assertEqual(expr.value, true_val) def test_wraps(self) -> None: """Test wrap classes.""" x = cp.Variable(2) expr = nonneg_wrap(x) assert expr.is_nonneg() expr = nonpos_wrap(x) assert expr.is_nonpos() Z = cp.Variable((2, 2)) U = cp.Variable((2, 2), complex=True) expr = psd_wrap(Z) assert expr.is_psd() assert not expr.is_complex() assert expr.is_symmetric() assert expr.is_hermitian() expr = psd_wrap(U) assert expr.is_psd() assert expr.is_complex() assert not expr.is_symmetric() assert expr.is_hermitian() expr = symmetric_wrap(Z) assert expr.is_symmetric() assert expr.is_hermitian() expr = skew_symmetric_wrap(Z) assert expr.is_skew_symmetric() expr = hermitian_wrap(U) assert expr.is_hermitian() def test_expr_does_not_support_cpp_warning(self): from cvxpy.atoms.affine.sum import Sum class SumNotSupportedInCPP(Sum): def _supports_cpp(self): return False x = Variable(2) prob = Problem(Minimize(0), [SumNotSupportedInCPP(x) == 1]) with pytest.warns( UserWarning, match="The problem includes expressions that don't support " "CPP backend. Defaulting to the SCIPY backend " "for canonicalization.", ): prob.solve() def test_expr_does_not_support_cpp_error(self): from cvxpy.atoms.affine.sum import Sum class SumNotSupportedInCPP(Sum): def _supports_cpp(self): return False x = Variable(2) prob = Problem(Minimize(0), [SumNotSupportedInCPP(x) == 1]) with pytest.raises( ValueError, match="The CPP backend cannot be used with problems " "that have expressions which do not support it", ): prob.solve(canon_backend=cp.CPP_CANON_BACKEND) class TestND_Expressions(): @pytest.fixture(autouse=True) def setup(self) -> None: self.x = Variable((2,2,2), name='x') self.target = (1+np.arange(8)).reshape(2,2,2) self.obj = cp.Minimize(0) def test_nd_variable(self) -> None: prob = cp.Problem(self.obj, [self.x == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(self.x.value, self.target) def test_nd_variable_warning(self) -> None: prob = cp.Problem(self.obj, [self.x == self.target]) warning_str = "The problem has an expression with dimension greater than 2. " \ "Defaulting to the SCIPY backend for canonicalization." with pytest.warns(UserWarning, match=warning_str): prob.solve() def test_nd_variable_value_error(self) -> None: prob = cp.Problem(self.obj, [self.x == self.target]) error_str = "Only the SCIPY and NUMPY backends are supported " \ "for problems with expressions of dimension greater than 2." with pytest.raises(ValueError, match=error_str): prob.solve(canon_backend=cp.CPP_CANON_BACKEND) def test_nd_mul_elem(self) -> None: expr = cp.multiply(self.x, 3) prob = cp.Problem(self.obj, [expr == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, self.target) def test_nd_div(self) -> None: expr = self.x / self.target prob = cp.Problem(self.obj, [expr == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, self.target) def test_nd_vstack(self) -> None: x = Variable((1, 2, 2)) z = Variable((1, 2, 2)) expr = cp.vstack([x,z]) prob = cp.Problem(self.obj, [expr == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, self.target) def test_nd_concatenate(self) -> None: x = cp.Variable((1, 2, 2)) z = cp.Variable((1, 2, 2)) expr = cp.concatenate([x,z], axis = 0) prob = cp.Problem(self.obj, [expr == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, self.target) def test_nd_sum_expr(self) -> None: x = [cp.Variable((2,2,2)) for _ in range(10)] expr = sum(x) prob = cp.Problem(self.obj, [expr == self.target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, self.target) @pytest.mark.parametrize("axis", [(0),(1),(2),((0,1)),((0,2)),((2,1))]) def test_nd_sum(self, axis) -> None: expr = cp.sum(self.x, axis=axis, keepdims=True) y = self.target.sum(axis=axis, keepdims=True) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("axis", [(0),(1),(2),((0,1)),((0,2)),((2,1))]) def test_nd_parametrized_sum(self, axis) -> None: param = cp.Parameter((2,2,2)) param.value = np.arange(8).reshape(2,2,2) expr = cp.multiply(self.x, param).sum(axis=axis) target = self.target.sum(axis=axis) prob = cp.Problem(self.obj, [expr == target]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, target) @pytest.mark.parametrize("axis", [(0,2,4,5),((4,5)),((0,2,3,1)),((5,3,1)), ((0,1,2,5))]) def test_nd_big_sum(self, axis) -> None: in_shape = (6,5,4,3,2,1) expr = cp.Variable(shape=in_shape).sum(axis=axis, keepdims=True) y = np.ones(in_shape).sum(axis=axis, keepdims=True) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @given(integer_array_indices(shape=(2,2,2))) def test_nd_integer_index(self, s) -> None: expr = self.x[s] y = self.target[s] prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @given(axis=basic_indices(shape=(2,2,2), allow_newaxis=True)) def test_nd_basic_index(self, axis) -> None: # Skip examples with 0-d output. TODO allow 0-d expressions in cvxpy. def is_zero_dim_output(axis): return 0 in self.target[axis].shape assume(is_zero_dim_output(axis) is False) expr = self.x[axis] y = self.target[axis] prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @given(axis=integer_array_indices(shape=(6,5,4,3,2,1))) def test_nd_big_index(self, axis) -> None: in_shape = (6,5,4,3,2,1) expr = cp.Variable(shape=in_shape)[axis] y = np.ones(in_shape)[axis] prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @given(axis=arrays(shape=(2,2,2), dtype=bool)) def test_nd_bool_index(self, axis) -> None: def is_zero_dim_output(axis): return 0 in self.target[axis].shape assume(is_zero_dim_output(axis) is False) expr = self.x[axis] y = self.target[axis] prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) def test_nd_index_sum(self) -> None: expr = self.x[:,:,0].sum(axis=0) y = self.target[:,:,0].sum(axis=0) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("order", ['C', 'F']) @pytest.mark.parametrize("shape", [(20, 2, 30), (300, 2, 2), (1, 24, 5, 10), (240, 5, 1)]) def test_nd_reshape(self, order, shape) -> None: var = cp.Variable((5, 24, 10)) target = np.arange(1200).reshape((5, 24, 10)) expr = cp.reshape(var, shape, order=order) y = target.reshape(shape, order=order) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) def test_nd_transpose(self) -> None: var = cp.Variable((5, 24, 10)) target = np.arange(1200).reshape((5, 24, 10)) expr = var.T y = target.T prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("axes", [(0, 2, 1), (2, 0, 1), (1, 2, 0), (1, 0, 2), (-1, 0, 1)]) def test_nd_transpose_axes(self, axes) -> None: var = cp.Variable((5, 24, 10)) target = np.arange(1200).reshape((5, 24, 10)) expr = cp.transpose(var, axes=axes) y = target.transpose(axes) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("axes", [(0, 2, 1, 3), (2, 3, 0, 1), (3, 1, 2, 0)]) def test_permute_dims(self, axes) -> None: var = cp.Variable((5, 6, 5, 6)) target = np.arange(900).reshape((5, 6, 5, 6)) expr = cp.permute_dims(var, axes=axes) y = target.transpose(axes) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("axis1, axis2", [(0, 1), (1, 3), (3, 2), (0, 3), (1, 2)]) def test_swapaxes(self, axis1, axis2) -> None: var = cp.Variable((5, 12, 6, 2)) target = np.arange(720).reshape((5, 12, 6, 2)) expr = cp.swapaxes(var, axis1=axis1, axis2=axis2) y = target.swapaxes(axis1, axis2) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("source, destination", [([0], [2]), ([0, 1], [3, 2]), ([0, 1, 2], [3, 2, 1])]) def test_moveaxis(self, source, destination) -> None: var = cp.Variable((5, 2, 6, 12)) target = np.arange(720).reshape((5, 2, 6, 12)) expr = cp.moveaxis(var, source=source, destination=destination) y = np.moveaxis(target, source, destination) prob = cp.Problem(self.obj, [expr == y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(expr.value, y) @pytest.mark.parametrize("shapes", [((3),(253, 253, 3)), ((7, 1, 5),(8, 7, 6, 5)), ((1),(5, 4)), ((4),(5, 4)), ((15, 1, 5), (15, 3, 5)), ((3, 5), (15, 3, 5)), ((3, 1), (15, 3, 5))]) def test_nd_broadcast(self, shapes) -> None: x = cp.Variable(shapes[0]) y = cp.broadcast_to(x, shape=shapes[1]) assert y.shape == shapes[1] prob = cp.Problem(cp.Minimize(cp.sum(y)), [y == 1]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(y.value, 1) @pytest.mark.parametrize("shapes", [((3), (2, 2, 2)), ((3), (4)), ((2, 1),(8, 4, 3))]) def test_nd_broadcast_error(self, shapes) -> None: error_str = "operands could not be broadcast together" with pytest.raises(Exception, match=error_str): x = cp.Variable(shapes[0]) y = cp.broadcast_to(x, shape=shapes[1]) assert y.shape is not shapes[1] @pytest.mark.parametrize("shapes", [((5),(3, 1)), ((15, 1),(8)), ((3), (2, 1))]) def test_no_segfault_multiply(self, shapes) -> None: """ This test ensures that no error is raised when multiplying two broadcastable array shapes <= 2. Previously this would cause a segfault in the CPP backend. """ x = cp.Variable(shapes[0]) target = np.arange(np.prod(shapes[0])).reshape(shapes[0]) a = np.arange(np.prod(shapes[1])).reshape(shapes[1]) b = np.arange(np.prod(shapes[1])).reshape(shapes[1]) obj = cp.sum(cp.max(cp.multiply(a, x) + b, axis=0)) prob = cp.Problem(cp.Minimize(obj), [x == target]) prob.solve() assert np.allclose(x.value, target) @pytest.mark.parametrize("shapes", [((3),(252, 253, 3)), ((7, 1, 5),(8, 7, 6, 5)), ((2, 1, 2), (2, 3, 2)), ((15, 1, 5), (15, 3, 5)), ((3, 5), (15, 3, 5)), ((3, 1), (15, 3, 5))]) def test_nd_multiply_broadcast(self, shapes) -> None: x = cp.Variable(shapes[0]) y = np.arange(np.prod(shapes[1])).reshape(shapes[1]) expr = cp.multiply(x, y) target = np.arange(np.prod(shapes[0])).reshape(shapes[0]) prob = cp.Problem(cp.Minimize(cp.sum(expr)), [expr == target * y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(x.value, target) @pytest.mark.parametrize("shapes", [((3),(252, 253, 3)), ((7, 1, 5),(8, 7, 6, 5)), ((2, 1, 2), (2, 3, 2)), ((3, 1), (15, 3, 5))]) def test_nd_add_broadcast(self, shapes) -> None: x = cp.Variable(shapes[0]) y = np.arange(np.prod(shapes[1])).reshape(shapes[1]) expr = x + y target = np.arange(np.prod(shapes[0])).reshape(shapes[0]) prob = cp.Problem(cp.Minimize(cp.sum(expr)), [expr == target + y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(x.value, target) @given(shape=broadcastable_shapes((8, 14, 8, 28), max_dims=4)) def test_nd_broadcast_generated(self, shape) -> None: x = cp.Variable((8, 14, 8, 28)) y = np.arange(np.prod((shape))).reshape((shape)) expr = x - y target = np.arange(np.prod((8,14,8,28))).reshape(8,14,8,28) prob = cp.Problem(cp.Minimize(cp.sum(expr)), [expr == target - y]) prob.solve(canon_backend=cp.SCIPY_CANON_BACKEND) assert np.allclose(x.value, target) #TODO make tests pass, support nd matmul def test_nd_matmul_exception(self) -> None: error_str = "Multiplication with N-d arrays is not yet supported" with pytest.raises(ValueError, match=error_str): x = cp.Variable((5,20,3)) y = cp.Variable((3,10)) x @ y #TODO make tests pass, support cumsum with multiple axes def test_nd_cumsum_warning(self) -> None: warning_str = "cumsum is only implemented for 1D or 2D arrays and might not" with pytest.raises(UserWarning, match=warning_str): x = cp.Variable((5,20,3)) cp.cumsum(x, axis=1)