Test failure with numpy 2.1

[bnavigator]

Running the test suite of 0.7.0 with numpy 2 installed (without numba):

[   28s] + pytest-3.10 --ignore=_build.python310 --ignore=_build.python312 --ignore=_build.python311 -v --ignore tests/test_einops.py --ignore tests/test_numba.py
[   29s] ============================= test session starts ==============================
[   29s] platform linux -- Python 3.10.14, pytest-8.3.2, pluggy-1.5.0 -- /usr/bin/python3.10
[   29s] cachedir: .pytest_cache
[   29s] hypothesis profile 'default' -> database=DirectoryBasedExampleDatabase(PosixPath('/home/abuild/rpmbuild/BUILD/xarray-einstats-0.7.0/.hypothesis/examples'))
[   29s] rootdir: /home/abuild/rpmbuild/BUILD/xarray-einstats-0.7.0
[   29s] configfile: pyproject.toml
[   29s] testpaths: tests
[   29s] plugins: hypothesis-6.111.1
[   29s] collecting ... collected 235 it

[   30s] =================================== FAILURES ===================================
[   30s] __________________________ test_linalg_accessor_solve __________________________
[   30s]
[   30s] hermitian = <xarray.DataArray (batch: 3, dim: 2, dim2: 2)> Size: 96B
[   30s] array([[[ 2. ,  1. ],
[   30s]         [ 1. ,  2. ]],
[   30s]
[   30s]        [[ 5. , ...
[   30s]         [ 3. ,  5. ]],
[   30s]
[   30s]        [[ 4. , -0.7],
[   30s]         [-0.7,  4. ]]])
[   30s] Dimensions without coordinates: batch, dim, dim2
[   30s]
[   30s]     def test_linalg_accessor_solve(hermitian):
[   30s]         db = hermitian.std("dim2")
[   30s] >       da_ac = hermitian.linalg.solve(db, dims=("dim", "dim2"))
[   30s]
[   30s] tests/test_accessors.py:75:
[   30s] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
[   30s] ../../BUILDROOT/python-xarray-einstats-0.7.0-0.x86_64/usr/lib/python3.10/site-packages/xarray_einstats/accessors.py:116: in solve
[   30s]     return solve(self._obj, db, dims=dims, **kwargs)
[   30s] ../../BUILDROOT/python-xarray-einstats-0.7.0-0.x86_64/usr/lib/python3.10/site-packages/xarray_einstats/linalg.py:722: in solve
[   30s]     return xr.apply_ufunc(
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:1268: in apply_ufunc
[   30s]     return apply_dataarray_vfunc(
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:312: in apply_dataarray_vfunc
[   30s]     result_var = func(*data_vars)
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:821: in apply_variable_ufunc
[   30s]     result_data = func(*input_data)
[   30s] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
[   30s]
[   30s] a = array([[[ 2. ,  1. ],
[   30s]         [ 1. ,  2. ]],
[   30s]
[   30s]        [[ 5. ,  3. ],
[   30s]         [ 3. ,  5. ]],
[   30s]
[   30s]        [[ 4. , -0.7],
[   30s]         [-0.7,  4. ]]])
[   30s] b = array([[0.5 , 0.5 ],
[   30s]        [1.  , 1.  ],
[   30s]        [2.35, 2.35]])
[   30s]
[   30s]     @array_function_dispatch(_solve_dispatcher)
[   30s]     def solve(a, b):
[   30s]         """
[   30s]         Solve a linear matrix equation, or system of linear scalar equations.
[   30s]
[   30s]         Computes the "exact" solution, `x`, of the well-determined, i.e., full
[   30s]         rank, linear matrix equation `ax = b`.
[   30s]
[   30s]         Parameters
[   30s]         ----------
[   30s]         a : (..., M, M) array_like
[   30s]             Coefficient matrix.
[   30s]         b : {(M,), (..., M, K)}, array_like
[   30s]             Ordinate or "dependent variable" values.
[   30s]
[   30s]         Returns
[   30s]         -------
[   30s]         x : {(..., M,), (..., M, K)} ndarray
[   30s]             Solution to the system a x = b.  Returned shape is (..., M) if b is
[   30s]             shape (M,) and (..., M, K) if b is (..., M, K), where the "..." part is
[   30s]             broadcasted between a and b.
[   30s]
[   30s]         Raises
[   30s]         ------
[   30s]         LinAlgError
[   30s]             If `a` is singular or not square.
[   30s]
[   30s]         See Also
[   30s]         --------
[   30s]         scipy.linalg.solve : Similar function in SciPy.
[   30s]
[   30s]         Notes
[   30s]         -----
[   30s]
[   30s]         .. versionadded:: 1.8.0
[   30s]
[   30s]         Broadcasting rules apply, see the `numpy.linalg` documentation for
[   30s]         details.
[   30s]
[   30s]         The solutions are computed using LAPACK routine ``_gesv``.
[   30s]
[   30s]         `a` must be square and of full-rank, i.e., all rows (or, equivalently,
[   30s]         columns) must be linearly independent; if either is not true, use
[   30s]         `lstsq` for the least-squares best "solution" of the
[   30s]         system/equation.
[   30s]
[   30s]         .. versionchanged:: 2.0
[   30s]
[   30s]            The b array is only treated as a shape (M,) column vector if it is
[   30s]            exactly 1-dimensional. In all other instances it is treated as a stack
[   30s]            of (M, K) matrices. Previously b would be treated as a stack of (M,)
[   30s]            vectors if b.ndim was equal to a.ndim - 1.
[   30s]
[   30s]         References
[   30s]         ----------
[   30s]         .. [1] G. Strang, *Linear Algebra and Its Applications*, 2nd Ed., Orlando,
[   30s]                FL, Academic Press, Inc., 1980, pg. 22.
[   30s]
[   30s]         Examples
[   30s]         --------
[   30s]         Solve the system of equations:
[   30s]         ``x0 + 2 * x1 = 1`` and
[   30s]         ``3 * x0 + 5 * x1 = 2``:
[   30s]
[   30s]         >>> import numpy as np
[   30s]         >>> a = np.array([[1, 2], [3, 5]])
[   30s]         >>> b = np.array([1, 2])
[   30s]         >>> x = np.linalg.solve(a, b)
[   30s]         >>> x
[   30s]         array([-1.,  1.])
[   30s]
[   30s]         Check that the solution is correct:
[   30s]
[   30s]         >>> np.allclose(np.dot(a, x), b)
[   30s]         True
[   30s]
[   30s]         """
[   30s]         a, _ = _makearray(a)
[   30s]         _assert_stacked_2d(a)
[   30s]         _assert_stacked_square(a)
[   30s]         b, wrap = _makearray(b)
[   30s]         t, result_t = _commonType(a, b)
[   30s]
[   30s]         # We use the b = (..., M,) logic, only if the number of extra dimensions
[   30s]         # match exactly
[   30s]         if b.ndim == 1:
[   30s]             gufunc = _umath_linalg.solve1
[   30s]         else:
[   30s]             gufunc = _umath_linalg.solve
[   30s]
[   30s]         signature = 'DD->D' if isComplexType(t) else 'dd->d'
[   30s]         with errstate(call=_raise_linalgerror_singular, invalid='call',
[   30s]                       over='ignore', divide='ignore', under='ignore'):
[   30s] >           r = gufunc(a, b, signature=signature)
[   30s] E           ValueError: solve: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (m,m),(m,n)->(m,n) (size 3 is different from 2)
[   30s]
[   30s] /usr/lib64/python3.10/site-packages/numpy/linalg/_linalg.py:413: ValueError
[   30s] ___________________________ TestWrappers.test_solve ____________________________
[   30s]
[   30s] self = <tests.test_linalg.TestWrappers object at 0x7f0217262710>
[   30s] matrices = <xarray.DataArray (batch: 10, experiment: 3, dim: 4, dim2: 4)> Size: 4kB
[   30s] 0.11 0.3897 1.4 2.2 0.3435 0.2578 0.4345 ... 0.3557 1.105 2.256 2.58 1.683 1.311
[   30s] Dimensions without coordinates: batch, experiment, dim, dim2
[   30s]
[   30s]     def test_solve(self, matrices):
[   30s]         b = matrices.std("dim2")
[   30s] >       y = solve(matrices, b, dims=("dim", "dim2"))
[   30s]
[   30s] tests/test_linalg.py:277:
[   30s] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
[   30s] ../../BUILDROOT/python-xarray-einstats-0.7.0-0.x86_64/usr/lib/python3.10/site-packages/xarray_einstats/linalg.py:722: in solve
[   30s]     return xr.apply_ufunc(
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:1268: in apply_ufunc
[   30s]     return apply_dataarray_vfunc(
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:312: in apply_dataarray_vfunc
[   30s]     result_var = func(*data_vars)
[   30s] /usr/lib/python3.10/site-packages/xarray/core/computation.py:821: in apply_variable_ufunc
[   30s]     result_data = func(*input_data)
[   30s] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
[   30s]
[   30s] a = array([[[[1.10014813e-01, 3.89656874e-01, 1.39954096e+00,
[   30s]           2.20014810e+00],
[   30s]          [3.43493819e-01, 2.57835...-01,
[   30s]           1.10546137e+00],
[   30s]          [2.25591889e+00, 2.57958002e+00, 1.68260005e+00,
[   30s]           1.31130995e+00]]]])
[   30s] b = array([[[0.83098024, 0.12226531, 1.27443529, 0.69294934],
[   30s]         [0.52431474, 0.48759741, 0.03113587, 0.9050297 ],
[   30s]   ...,
[   30s]         [1.17554304, 1.2982992 , 0.85877319, 0.15673694],
[   30s]         [0.68007424, 1.21224985, 0.35092172, 0.49223187]]])
[   30s]
[   30s]     @array_function_dispatch(_solve_dispatcher)
[   30s]     def solve(a, b):
[   30s]         """
[   30s]         Solve a linear matrix equation, or system of linear scalar equations.
[   30s]
[   30s]         Computes the "exact" solution, `x`, of the well-determined, i.e., full
[   30s]         rank, linear matrix equation `ax = b`.
[   30s]
[   30s]         Parameters
[   30s]         ----------
[   30s]         a : (..., M, M) array_like
[   30s]             Coefficient matrix.
[   30s]         b : {(M,), (..., M, K)}, array_like
[   30s]             Ordinate or "dependent variable" values.
[   30s]
[   30s]         Returns
[   30s]         -------
[   30s]         x : {(..., M,), (..., M, K)} ndarray
[   30s]             Solution to the system a x = b.  Returned shape is (..., M) if b is
[   30s]             shape (M,) and (..., M, K) if b is (..., M, K), where the "..." part is
[   30s]             broadcasted between a and b.
[   30s]
[   30s]         Raises
[   30s]         ------
[   30s]         LinAlgError
[   30s]             If `a` is singular or not square.
[   30s]
[   30s]         See Also
[   30s]         --------
[   30s]         scipy.linalg.solve : Similar function in SciPy.
[   30s]
[   30s]         Notes
[   30s]         -----
[   30s]
[   30s]         .. versionadded:: 1.8.0
[   30s]
[   30s]         Broadcasting rules apply, see the `numpy.linalg` documentation for
[   30s]         details.
[   30s]
[   30s]         The solutions are computed using LAPACK routine ``_gesv``.
[   30s]
[   30s]         `a` must be square and of full-rank, i.e., all rows (or, equivalently,
[   30s]         columns) must be linearly independent; if either is not true, use
[   30s]         `lstsq` for the least-squares best "solution" of the
[   30s]         system/equation.
[   30s]
[   30s]         .. versionchanged:: 2.0
[   30s]
[   30s]            The b array is only treated as a shape (M,) column vector if it is
[   30s]            exactly 1-dimensional. In all other instances it is treated as a stack
[   30s]            of (M, K) matrices. Previously b would be treated as a stack of (M,)
[   30s]            vectors if b.ndim was equal to a.ndim - 1.
[   30s]
[   30s]         References
[   30s]         ----------
[   30s]         .. [1] G. Strang, *Linear Algebra and Its Applications*, 2nd Ed., Orlando,
[   30s]                FL, Academic Press, Inc., 1980, pg. 22.
[   30s]
[   30s]         Examples
[   30s]         --------
[   30s]         Solve the system of equations:
[   30s]         ``x0 + 2 * x1 = 1`` and
[   30s]         ``3 * x0 + 5 * x1 = 2``:
[   30s]
[   30s]         >>> import numpy as np
[   30s]         >>> a = np.array([[1, 2], [3, 5]])
[   30s]         >>> b = np.array([1, 2])
[   30s]         >>> x = np.linalg.solve(a, b)
[   30s]         >>> x
[   30s]         array([-1.,  1.])
[   30s]
[   30s]         Check that the solution is correct:
[   30s]
[   30s]         >>> np.allclose(np.dot(a, x), b)
[   30s]         True
[   30s]
[   30s]         """
[   30s]         a, _ = _makearray(a)
[   30s]         _assert_stacked_2d(a)
[   30s]         _assert_stacked_square(a)
[   30s]         b, wrap = _makearray(b)
[   30s]         t, result_t = _commonType(a, b)
[   30s]
[   30s]         # We use the b = (..., M,) logic, only if the number of extra dimensions
[   30s]         # match exactly
[   30s]         if b.ndim == 1:
[   30s]             gufunc = _umath_linalg.solve1
[   30s]         else:
[   30s]             gufunc = _umath_linalg.solve
[   30s]
[   30s]         signature = 'DD->D' if isComplexType(t) else 'dd->d'
[   30s]         with errstate(call=_raise_linalgerror_singular, invalid='call',
[   30s]                       over='ignore', divide='ignore', under='ignore'):
[   30s] >           r = gufunc(a, b, signature=signature)
[   30s] E           ValueError: solve: Input operand 1 has a mismatch in its core dimension 0, with gufunc signature (m,m),(m,n)->(m,n) (size 3 is different from 4)
[   30s]
[   30s] /usr/lib64/python3.10/site-packages/numpy/linalg/_linalg.py:413: ValueError
[   30s] =============================== warnings summary ===============================
[   30s] tests/test_stats.py::TestMvNormal::test_rvs_method[vals0]
[   30s] tests/test_stats.py::TestMvNormal::test_rvs_method[vals1]
[   30s]   /home/abuild/rpmbuild/BUILDROOT/python-xarray-einstats-0.7.0-0.x86_64/usr/lib/python3.10/site-packages/xarray_einstats/stats.py:384: PendingDeprecationWarning: The `dims` argument has been renamed to `dim`, and will be removed in the future. This renaming is taking place throughout xarray over the next few releases.
[   30s]     return mean + xr.dot(cov_chol, samples, dims=dim2)
[   30s]
[   30s] tests/test_stats.py::TestMvNormal::test_pdf_method[vals0]
[   30s] tests/test_stats.py::TestMvNormal::test_pdf_method[vals1]
[   30s]   /home/abuild/rpmbuild/BUILDROOT/python-xarray-einstats-0.7.0-0.x86_64/usr/lib/python3.10/site-packages/xarray_einstats/stats.py:397: PendingDeprecationWarning: The `dims` argument has been renamed to `dim`, and will be removed in the future. This renaming is taking place throughout xarray over the next few releases.
[   30s]     maha = np.square(xr.dot(x_mu.rename({dim1: dim2}), u_mat, dims=dim2)).sum(dim=dim1)
[   30s]
[   30s] -- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
[   30s] =========================== short test summary info ============================
[   30s] FAILED tests/test_accessors.py::test_linalg_accessor_solve - ValueError: solv...
[   30s] FAILED tests/test_linalg.py::TestWrappers::test_solve - ValueError: solve: In...
[   30s] ============= 2 failed, 230 passed, 3 skipped, 4 warnings in 1.59s =============

[OriolAbril]

Thanks for reporting, will try to fix asap.