MAINT: 1.11.4 backports

.circleci/config.yml

@@ -5,7 +5,7 @@ _defaults: &defaults
   docker:
     # CircleCI maintains a library of pre-built images
     # documented at https://circleci.com/docs/2.0/circleci-images/
-    - image: cimg/python:3.9
+    - image: cimg/python:3.11
   working_directory: ~/repo
 
 commands:
@@ -77,6 +77,7 @@ jobs:
             pip install cython
             pip install numpy==1.23.5
             pip install -r doc_requirements.txt
+            pip install "myst-nb<1.0.0"
             # `asv` pin because of slowdowns reported in gh-15568
             pip install mpmath gmpy2 "asv==0.4.2" pythran ninja meson click rich-click doit pydevtool pooch
             pip install pybind11
@@ -118,7 +119,7 @@ jobs:
           name: build docs
           no_output_timeout: 25m
           command: |
-            export PYTHONPATH=$PWD/build-install/lib/python3.9/site-packages
+            export PYTHONPATH=$PWD/build-install/lib/python3.11/site-packages
             python dev.py --no-build doc -j2
 
       - store_artifacts:
@@ -145,7 +146,7 @@ jobs:
           name: run asv
           no_output_timeout: 30m
           command: |
-            export PYTHONPATH=$PWD/build-install/lib/python3.9/site-packages
+            export PYTHONPATH=$PWD/build-install/lib/python3.11/site-packages
             cd benchmarks
             asv machine --machine CircleCI
             export SCIPY_GLOBAL_BENCH_NUMTRIALS=1
@@ -173,7 +174,7 @@ jobs:
           no_output_timeout: 25m
           command: |
             sudo apt-get install -y wamerican-small
-            export PYTHONPATH=$PWD/build-install/lib/python3.9/site-packages
+            export PYTHONPATH=$PWD/build-install/lib/python3.11/site-packages
             python dev.py --no-build refguide-check
 
 # Upload build output to scipy/devdocs repository, using SSH deploy keys.

doc/source/conf.py

@@ -183,6 +183,12 @@
     message=r'There is no current event loop',
     category=DeprecationWarning,
 )
+# TODO: remove after gh-19228 resolved:
+warnings.filterwarnings(
+    'ignore',
+    message=r'.*path is deprecated.*',
+    category=DeprecationWarning,
+)
 
 # -----------------------------------------------------------------------------
 # HTML output

doc/source/release/1.11.4-notes.rst

@@ -12,12 +12,51 @@ compared to 1.11.3.
 Authors
 =======
 * Name (commits)
+* Jake Bowhay (2)
+* Ralf Gommers (4)
+* Julien Jerphanion (2)
+* Nikolay Mayorov (2)
+* Melissa Weber Mendonça (1)
+* Tirth Patel (1)
+* Tyler Reddy (22)
+* Dan Schult (3)
+* Nicolas Vetsch (1) +
+
+A total of 9 people contributed to this release.
+People with a "+" by their names contributed a patch for the first time.
+This list of names is automatically generated, and may not be fully complete.
+
 
 
 Issues closed for 1.11.4
 ------------------------
 
+* `#19189 <https://github.com/scipy/scipy/issues/19189>`__: Contradiction in \`pyproject.toml\` requirements?
+* `#19228 <https://github.com/scipy/scipy/issues/19228>`__: Doc build fails with Python 3.11
+* `#19245 <https://github.com/scipy/scipy/issues/19245>`__: BUG: upcasting of indices dtype from DIA to COO/CSR/BSR arrays
+* `#19351 <https://github.com/scipy/scipy/issues/19351>`__: BUG: Regression in 1.11.3 can still fail for \`optimize.least_squares\`...
+* `#19357 <https://github.com/scipy/scipy/issues/19357>`__: BUG: build failure with Xcode 15 linker
+* `#19359 <https://github.com/scipy/scipy/issues/19359>`__: BUG: DiscreteAliasUrn construction fails with UNURANError for...
+* `#19387 <https://github.com/scipy/scipy/issues/19387>`__: BUG: problem importing libgfortran.5.dylib on macOS Sonoma
+* `#19403 <https://github.com/scipy/scipy/issues/19403>`__: BUG: scipy.sparse.lil_matrix division by complex number leads...
+* `#19437 <https://github.com/scipy/scipy/issues/19437>`__: BUG: can't install scipy on mac m1 with poetry due to incompatible...
+* `#19500 <https://github.com/scipy/scipy/issues/19500>`__: DOC: doc build failing
+* `#19513 <https://github.com/scipy/scipy/issues/19513>`__: BUG: Python version constraints in releases causes issues for...
 
 
 Pull requests for 1.11.4
 ------------------------
+
+* `#19230 <https://github.com/scipy/scipy/pull/19230>`__: DOC, MAINT: workaround for py311 docs
+* `#19307 <https://github.com/scipy/scipy/pull/19307>`__: set idx_dtype in sparse dia_array.tocoo
+* `#19316 <https://github.com/scipy/scipy/pull/19316>`__: MAINT: Prep 1.11.4
+* `#19320 <https://github.com/scipy/scipy/pull/19320>`__: BLD: fix up version parsing issue in cythonize.py for setup.py...
+* `#19329 <https://github.com/scipy/scipy/pull/19329>`__: DOC: stats.chisquare: result object contains attribute 'statistic'
+* `#19335 <https://github.com/scipy/scipy/pull/19335>`__: BUG: fix pow method for sparrays with power zero
+* `#19364 <https://github.com/scipy/scipy/pull/19364>`__: MAINT, BUG: stats: update the UNU.RAN submodule with DAU fix
+* `#19379 <https://github.com/scipy/scipy/pull/19379>`__: BUG: Restore the original behavior of 'trf' from least_squares...
+* `#19400 <https://github.com/scipy/scipy/pull/19400>`__: BLD: use classic linker on macOS 14 (Sonoma), the new linker...
+* `#19408 <https://github.com/scipy/scipy/pull/19408>`__: BUG: Fix typecasting problem in scipy.sparse.lil_matrix truediv
+* `#19504 <https://github.com/scipy/scipy/pull/19504>`__: DOC, MAINT: Bump CircleCI Python version to 3.11
+* `#19517 <https://github.com/scipy/scipy/pull/19517>`__: MAINT, REL: unpin Python 1.11.x branch
+* `#19550 <https://github.com/scipy/scipy/pull/19550>`__: MAINT, BLD: poetry loongarch shims

meson.build

@@ -73,9 +73,7 @@ endif
 if host_machine.system() == 'os400'
   # IBM i system, needed to avoid build errors - see gh-17193
   add_project_arguments('-D__STDC_FORMAT_MACROS', language : 'cpp')
-  add_project_link_arguments('-Wl,-bnotextro', language : 'c')
-  add_project_link_arguments('-Wl,-bnotextro', language : 'cpp')
-  add_project_link_arguments('-Wl,-bnotextro', language : 'fortran')
+  add_project_link_arguments('-Wl,-bnotextro', language : ['c', 'cpp', 'fortran'])
 endif
 
 # Adding at project level causes many spurious -lgfortran flags.
@@ -85,6 +83,11 @@ if ff.has_argument('-Wno-conversion')
   add_project_arguments('-Wno-conversion', language: 'fortran')
 endif
 
+if host_machine.system() == 'darwin' and cc.has_link_argument('-Wl,-ld_classic')
+  # New linker introduced in macOS 14 not working yet, see gh-19357 and gh-19387
+  add_project_link_arguments('-Wl,-ld_classic', language : ['c', 'cpp', 'fortran'])
+endif
+
 # Intel compilers default to fast-math, so disable it if we detect Intel
 # compilers. A word of warning: this may not work with the conda-forge
 # compilers, because those have the annoying habit of including lots of flags

scipy/optimize/_lsq/least_squares.py

@@ -268,7 +268,9 @@ def least_squares(
         arguments, as shown at the end of the Examples section.
     x0 : array_like with shape (n,) or float
         Initial guess on independent variables. If float, it will be treated
-        as a 1-D array with one element.
+        as a 1-D array with one element. When `method` is 'trf', the initial
+        guess might be slightly adjusted to lie sufficiently within the given
+        `bounds`.
     jac : {'2-point', '3-point', 'cs', callable}, optional
         Method of computing the Jacobian matrix (an m-by-n matrix, where
         element (i, j) is the partial derivative of f[i] with respect to
@@ -822,7 +824,7 @@ def least_squares(
     ftol, xtol, gtol = check_tolerance(ftol, xtol, gtol, method)
 
     if method == 'trf':
-        x0 = make_strictly_feasible(x0, lb, ub, rstep=0)
+        x0 = make_strictly_feasible(x0, lb, ub)
 
     def fun_wrapped(x):
         return np.atleast_1d(fun(x, *args, **kwargs))

scipy/optimize/tests/test_least_squares.py

@@ -10,7 +10,7 @@
 
 from scipy.optimize import least_squares, Bounds
 from scipy.optimize._lsq.least_squares import IMPLEMENTED_LOSSES
-from scipy.optimize._lsq.common import EPS, make_strictly_feasible
+from scipy.optimize._lsq.common import EPS, make_strictly_feasible, CL_scaling_vector
 
 
 def fun_trivial(x, a=0):
@@ -811,17 +811,30 @@ def err(p, x, y):
     assert_allclose(res.x, np.array([0.4082241, 0.15530563]), atol=5e-5)
 
 
-def test_gh_18793():
+def test_gh_18793_and_19351():
     answer = 1e-12
     initial_guess = 1.1e-12
 
     def chi2(x):
         return (x-answer)**2
 
-    res = least_squares(chi2, x0=initial_guess, bounds=(0, np.inf))
+    gtol = 1e-15
+    res = least_squares(chi2, x0=initial_guess, gtol=1e-15, bounds=(0, np.inf))
+    # Original motivation: gh-18793
     # if we choose an initial condition that is close to the solution
     # we shouldn't return an answer that is further away from the solution
-    assert_allclose(res.x, answer, atol=initial_guess-answer)
+
+    # Update: gh-19351
+    # However this requirement does not go well with 'trf' algorithm logic.
+    # Some regressions were reported after the presumed fix.
+    # The returned solution is good as long as it satisfies the convergence
+    # conditions.
+    # Specifically in this case the scaled gradient will be sufficiently low.
+
+    scaling, _ = CL_scaling_vector(res.x, res.grad,
+                                   np.atleast_1d(0), np.atleast_1d(np.inf))
+    assert res.status == 1  # Converged by gradient
+    assert np.linalg.norm(res.grad * scaling, ord=np.inf) < gtol
 
 
 def test_gh_19103():

scipy/sparse/_data.py

@@ -106,12 +106,25 @@ def power(self, n, dtype=None):
 
         Parameters
         ----------
-        n : n is a scalar
+        n : scalar
+            n is a non-zero scalar (nonzero avoids dense ones creation)
+            If zero power is desired, special case it to use `np.ones`
 
         dtype : If dtype is not specified, the current dtype will be preserved.
+
+        Raises
+        ------
+        NotImplementedError : if n is a zero scalar
+            If zero power is desired, special case it to use
+            `np.ones(A.shape, dtype=A.dtype)`
         """
         if not isscalarlike(n):
             raise NotImplementedError("input is not scalar")
+        if not n:
+            raise NotImplementedError(
+                "zero power is not supported as it would densify the matrix.\n"
+                "Use `np.ones(A.shape, dtype=A.dtype)` for this case."
+            )
 
         data = self._deduped_data()
         if dtype is not None:

scipy/sparse/_dia.py

@@ -157,7 +157,7 @@ def __init__(self, arg1, shape=None, dtype=None, copy=False):
             raise ValueError('offset array contains duplicate values')
 
     def __repr__(self):
-        format = _formats[self.getformat()][1]
+        format = _formats[self.format][1]
         return "<%dx%d sparse matrix of type '%s'\n" \
                "\twith %d stored elements (%d diagonals) in %s format>" % \
                (self.shape + (self.dtype.type, self.nnz, self.data.shape[0],
@@ -402,6 +402,11 @@ def tocoo(self, copy=False):
         mask &= (self.data != 0)
         row = row[mask]
         col = np.tile(offset_inds, num_offsets)[mask.ravel()]
+        idx_dtype = self._get_index_dtype(
+            arrays=(self.offsets,), maxval=max(self.shape)
+        )
+        row = row.astype(idx_dtype, copy=False)
+        col = col.astype(idx_dtype, copy=False)
         data = self.data[mask]
         # Note: this cannot set has_canonical_format=True, because despite the
         # lack of duplicates, we do not generate sorted indices.

scipy/sparse/_lil.py

@@ -348,6 +348,7 @@ def _mul_scalar(self, other):
     def __truediv__(self, other):           # self / other
         if isscalarlike(other):
             new = self.copy()
+            new.dtype = np.result_type(self, other)
             # Divide every element by this scalar
             for j, rowvals in enumerate(new.data):
                 new.data[j] = [val/other for val in rowvals]

scipy/sparse/tests/test_array_api.py

@@ -138,10 +138,16 @@ def test_matmul(A):
     assert np.all((A @ A.T).todense() == A.dot(A.T).todense())
 
 
-@parametrize_square_sparrays
-def test_pow(B):
-    assert (B**0)._is_array, "Expected array, got matrix"
-    assert (B**2)._is_array, "Expected array, got matrix"
+@parametrize_sparrays
+def test_power_operator(A):
+    assert isinstance((A**2), scipy.sparse.sparray), "Expected array, got matrix"
+
+    # https://github.com/scipy/scipy/issues/15948
+    npt.assert_equal((A**2).todense(), (A.todense())**2)
+
+    # power of zero is all ones (dense) so helpful msg exception
+    with pytest.raises(NotImplementedError, match="zero power"):
+        A**0
 
 
 @parametrize_sparrays
@@ -344,12 +350,6 @@ def test_spilu():
     npt.assert_allclose(LU.solve(np.array([1, 2, 3, 4])), [1, 0, 0, 0])
 
 
-@parametrize_sparrays
-def test_power_operator(A):
-    # https://github.com/scipy/scipy/issues/15948
-    npt.assert_equal((A**2).todense(), (A.todense())**2)
-
-
 @pytest.mark.parametrize(
     "cls,indices_attrs",
     [

scipy/sparse/tests/test_base.py

@@ -3842,7 +3842,7 @@ def test_has_sorted_indices(self):
         indptr = np.array([0, 2])
         M = csr_matrix((data, sorted_inds, indptr)).copy()
         assert_equal(True, M.has_sorted_indices)
-        assert type(M.has_sorted_indices) == bool
+        assert isinstance(M.has_sorted_indices, bool)
 
         M = csr_matrix((data, unsorted_inds, indptr)).copy()
         assert_equal(False, M.has_sorted_indices)
@@ -3874,7 +3874,7 @@ def test_has_canonical_format(self):
 
         M = csr_matrix((data, indices, indptr)).copy()
         assert_equal(False, M.has_canonical_format)
-        assert type(M.has_canonical_format) == bool
+        assert isinstance(M.has_canonical_format, bool)
 
         # set by deduplicating
         M.sum_duplicates()
@@ -4204,6 +4204,14 @@ def test_scalar_mul(self):
         x = x*0
         assert_equal(x[0, 0], 0)
 
+    def test_truediv_scalar(self):
+        A = self.spcreator((3, 2))
+        A[0, 1] = -10
+        A[2, 0] = 20
+
+        assert_array_equal((A / 1j).toarray(), A.toarray() / 1j)
+        assert_array_equal((A / 9).toarray(), A.toarray() / 9)
+
     def test_inplace_ops(self):
         A = lil_matrix([[0, 2, 3], [4, 0, 6]])
         B = lil_matrix([[0, 1, 0], [0, 2, 3]])
@@ -4465,6 +4473,19 @@ def test_tocoo_gh10050(self):
         inds_are_sorted = np.all(np.diff(flat_inds) > 0)
         assert m.has_canonical_format == inds_are_sorted
 
+    def test_tocoo_tocsr_tocsc_gh19245(self):
+        # test index_dtype with tocoo, tocsr, tocsc
+        data = np.array([[1, 2, 3, 4]]).repeat(3, axis=0)
+        offsets = np.array([0, -1, 2], dtype=np.int32)
+        dia = sparse.dia_array((data, offsets), shape=(4, 4))
+
+        coo = dia.tocoo()
+        assert coo.col.dtype == np.int32
+        csr = dia.tocsr()
+        assert csr.indices.dtype == np.int32
+        csc = dia.tocsc()
+        assert csc.indices.dtype == np.int32
+
 
 TestDIA.init_class()
 
@@ -4918,6 +4939,7 @@ def cases_64bit():
         'test_large_dimensions_reshape': 'test actually requires 64-bit to work',
         'test_constructor_smallcol': 'test verifies int32 indexes',
         'test_constructor_largecol': 'test verifies int64 indexes',
+        'test_tocoo_tocsr_tocsc_gh19245': 'test verifies int32 indexes',
     }
 
     for cls in TEST_CLASSES:

scipy/stats/tests/test_sampling.py

@@ -292,24 +292,24 @@ def test_with_scipy_distribution():
     check_discr_samples(rng, pv, dist.stats())
 
 
-def check_cont_samples(rng, dist, mv_ex):
+def check_cont_samples(rng, dist, mv_ex, rtol=1e-7, atol=1e-1):
     rvs = rng.rvs(100000)
     mv = rvs.mean(), rvs.var()
     # test the moments only if the variance is finite
     if np.isfinite(mv_ex[1]):
-        assert_allclose(mv, mv_ex, rtol=1e-7, atol=1e-1)
+        assert_allclose(mv, mv_ex, rtol=rtol, atol=atol)
     # Cramer Von Mises test for goodness-of-fit
     rvs = rng.rvs(500)
     dist.cdf = np.vectorize(dist.cdf)
     pval = cramervonmises(rvs, dist.cdf).pvalue
     assert pval > 0.1
 
 
-def check_discr_samples(rng, pv, mv_ex):
+def check_discr_samples(rng, pv, mv_ex, rtol=1e-3, atol=1e-1):
     rvs = rng.rvs(100000)
     # test if the first few moments match
     mv = rvs.mean(), rvs.var()
-    assert_allclose(mv, mv_ex, rtol=1e-3, atol=1e-1)
+    assert_allclose(mv, mv_ex, rtol=rtol, atol=atol)
     # normalize
     pv = pv / pv.sum()
     # chi-squared test for goodness-of-fit
@@ -737,6 +737,13 @@ def pmf(self, x):
         with pytest.raises(ValueError, match=msg):
             DiscreteAliasUrn(dist)
 
+    def test_gh19359(self):
+        pv = special.softmax(np.ones((1533,)))
+        rng = DiscreteAliasUrn(pv, random_state=42)
+        # check the correctness
+        check_discr_samples(rng, pv, (1532 / 2, (1532**2 - 1) / 12),
+                            rtol=5e-3)
+
 
 class TestNumericalInversePolynomial:
     # Simple Custom Distribution