import pytest import numpy as np import scipy.sparse as sp import scipy.sparse.linalg as splin from numpy.testing import assert_allclose, assert_equal try: import sparse except Exception: sparse = None pytestmark = pytest.mark.skipif(sparse is None, reason="pydata/sparse not installed") msg = "pydata/sparse (0.15.1) does not implement necessary operations" sparse_params = (pytest.param("COO"), pytest.param("DOK", marks=[pytest.mark.xfail(reason=msg)])) scipy_sparse_classes = [ sp.bsr_matrix, sp.csr_matrix, sp.coo_matrix, sp.csc_matrix, sp.dia_matrix, sp.dok_matrix ] @pytest.fixture(params=sparse_params) def sparse_cls(request): return getattr(sparse, request.param) @pytest.fixture(params=scipy_sparse_classes) def sp_sparse_cls(request): return request.param @pytest.fixture def same_matrix(sparse_cls, sp_sparse_cls): np.random.seed(1234) A_dense = np.random.rand(9, 9) return sp_sparse_cls(A_dense), sparse_cls(A_dense) @pytest.fixture def matrices(sparse_cls): np.random.seed(1234) A_dense = np.random.rand(9, 9) A_dense = A_dense @ A_dense.T A_sparse = sparse_cls(A_dense) b = np.random.rand(9) return A_dense, A_sparse, b def test_isolve_gmres(matrices): # Several of the iterative solvers use the same # isolve.utils.make_system wrapper code, so test just one of them. A_dense, A_sparse, b = matrices x, info = splin.gmres(A_sparse, b, atol=1e-15) assert info == 0 assert isinstance(x, np.ndarray) assert_allclose(A_sparse @ x, b) def test_lsmr(matrices): A_dense, A_sparse, b = matrices res0 = splin.lsmr(A_dense, b) res = splin.lsmr(A_sparse, b) assert_allclose(res[0], res0[0], atol=1e-3) # test issue 17012 def test_lsmr_output_shape(): x = splin.lsmr(A=np.ones((10, 1)), b=np.zeros(10), x0=np.ones(1))[0] assert_equal(x.shape, (1,)) def test_lsqr(matrices): A_dense, A_sparse, b = matrices res0 = splin.lsqr(A_dense, b) res = splin.lsqr(A_sparse, b) assert_allclose(res[0], res0[0], atol=1e-5) def test_eigs(matrices): A_dense, A_sparse, v0 = matrices M_dense = np.diag(v0**2) M_sparse = A_sparse.__class__(M_dense) w_dense, v_dense = splin.eigs(A_dense, k=3, v0=v0) w, v = splin.eigs(A_sparse, k=3, v0=v0) assert_allclose(w, w_dense) assert_allclose(v, v_dense) for M in [M_sparse, M_dense]: w_dense, v_dense = splin.eigs(A_dense, M=M_dense, k=3, v0=v0) w, v = splin.eigs(A_sparse, M=M, k=3, v0=v0) assert_allclose(w, w_dense) assert_allclose(v, v_dense) w_dense, v_dense = splin.eigsh(A_dense, M=M_dense, k=3, v0=v0) w, v = splin.eigsh(A_sparse, M=M, k=3, v0=v0) assert_allclose(w, w_dense) assert_allclose(v, v_dense) def test_svds(matrices): A_dense, A_sparse, v0 = matrices u0, s0, vt0 = splin.svds(A_dense, k=2, v0=v0) u, s, vt = splin.svds(A_sparse, k=2, v0=v0) assert_allclose(s, s0) assert_allclose(np.abs(u), np.abs(u0)) assert_allclose(np.abs(vt), np.abs(vt0)) def test_lobpcg(matrices): A_dense, A_sparse, x = matrices X = x[:,None] w_dense, v_dense = splin.lobpcg(A_dense, X) w, v = splin.lobpcg(A_sparse, X) assert_allclose(w, w_dense) assert_allclose(v, v_dense) def test_spsolve(matrices): A_dense, A_sparse, b = matrices b2 = np.random.rand(len(b), 3) x0 = splin.spsolve(sp.csc_matrix(A_dense), b) x = splin.spsolve(A_sparse, b) assert isinstance(x, np.ndarray) assert_allclose(x, x0) x0 = splin.spsolve(sp.csc_matrix(A_dense), b) x = splin.spsolve(A_sparse, b, use_umfpack=True) assert isinstance(x, np.ndarray) assert_allclose(x, x0) x0 = splin.spsolve(sp.csc_matrix(A_dense), b2) x = splin.spsolve(A_sparse, b2) assert isinstance(x, np.ndarray) assert_allclose(x, x0) x0 = splin.spsolve(sp.csc_matrix(A_dense), sp.csc_matrix(A_dense)) x = splin.spsolve(A_sparse, A_sparse) assert isinstance(x, type(A_sparse)) assert_allclose(x.todense(), x0.todense()) def test_splu(matrices): A_dense, A_sparse, b = matrices n = len(b) sparse_cls = type(A_sparse) lu = splin.splu(A_sparse) assert isinstance(lu.L, sparse_cls) assert isinstance(lu.U, sparse_cls) _Pr_scipy = sp.csc_matrix((np.ones(n), (lu.perm_r, np.arange(n)))) _Pc_scipy = sp.csc_matrix((np.ones(n), (np.arange(n), lu.perm_c))) Pr = sparse_cls.from_scipy_sparse(_Pr_scipy) Pc = sparse_cls.from_scipy_sparse(_Pc_scipy) A2 = Pr.T @ lu.L @ lu.U @ Pc.T assert_allclose(A2.todense(), A_sparse.todense()) z = lu.solve(A_sparse.todense()) assert_allclose(z, np.eye(n), atol=1e-10) def test_spilu(matrices): A_dense, A_sparse, b = matrices sparse_cls = type(A_sparse) lu = splin.spilu(A_sparse) assert isinstance(lu.L, sparse_cls) assert isinstance(lu.U, sparse_cls) z = lu.solve(A_sparse.todense()) assert_allclose(z, np.eye(len(b)), atol=1e-3) def test_spsolve_triangular(matrices): A_dense, A_sparse, b = matrices A_sparse = sparse.tril(A_sparse) x = splin.spsolve_triangular(A_sparse, b) assert_allclose(A_sparse @ x, b) def test_onenormest(matrices): A_dense, A_sparse, b = matrices est0 = splin.onenormest(A_dense) est = splin.onenormest(A_sparse) assert_allclose(est, est0) def test_inv(matrices): A_dense, A_sparse, b = matrices x0 = splin.inv(sp.csc_matrix(A_dense)) x = splin.inv(A_sparse) assert_allclose(x.todense(), x0.todense()) def test_expm(matrices): A_dense, A_sparse, b = matrices x0 = splin.expm(sp.csc_matrix(A_dense)) x = splin.expm(A_sparse) assert_allclose(x.todense(), x0.todense()) def test_expm_multiply(matrices): A_dense, A_sparse, b = matrices x0 = splin.expm_multiply(A_dense, b) x = splin.expm_multiply(A_sparse, b) assert_allclose(x, x0) def test_eq(same_matrix): sp_sparse, pd_sparse = same_matrix assert (sp_sparse == pd_sparse).all() def test_ne(same_matrix): sp_sparse, pd_sparse = same_matrix assert not (sp_sparse != pd_sparse).any()