NumPy 2.0 also moved function_base

Handle warning in test_bounds_gauss2d_lsq
that is flaky and we do not really care for, so just ignore it.

astropy/modeling/tests/test_constraints.py

@@ -15,7 +15,6 @@
 from astropy.modeling.core import Fittable1DModel
 from astropy.modeling.parameters import Parameter
 from astropy.utils import minversion
-from astropy.utils.compat.numpycompat import NUMPY_LT_2_0
 from astropy.utils.compat.optional_deps import HAS_SCIPY
 from astropy.utils.exceptions import AstropyUserWarning
 
@@ -184,6 +183,7 @@ def test_bounds_slsqp(self):
         assert intercept + 10**-5 >= bounds["intercept"][0]
         assert intercept - 10**-5 <= bounds["intercept"][1]
 
+    @pytest.mark.filterwarnings("ignore:The fit may be unsuccessful")
     @pytest.mark.parametrize("fitter", fitters)
     def test_bounds_gauss2d_lsq(self, fitter):
         fitter = fitter()
@@ -204,21 +204,12 @@ def test_bounds_gauss2d_lsq(self, fitter):
             theta=0.5,
             bounds=bounds,
         )
-        if isinstance(fitter, (fitting.LevMarLSQFitter, fitting.DogBoxLSQFitter)):
-            with pytest.warns(AstropyUserWarning, match="The fit may be unsuccessful"):
-                model = fitter(gauss, X, Y, self.data)
+        if isinstance(fitter, fitting.TRFLSQFitter):
+            ctx = np.errstate(invalid="ignore", divide="ignore")
         else:
-            ctx2 = nullcontext()
-            if isinstance(fitter, fitting.TRFLSQFitter):
-                ctx = np.errstate(invalid="ignore", divide="ignore")
-                if not NUMPY_LT_2_0 or not SCIPY_LT_1_11_2:
-                    ctx2 = pytest.warns(
-                        AstropyUserWarning, match="The fit may be unsuccessful"
-                    )
-            else:
-                ctx = nullcontext()
-            with ctx, ctx2:
-                model = fitter(gauss, X, Y, self.data)
+            ctx = nullcontext()
+        with ctx:
+            model = fitter(gauss, X, Y, self.data)
         x_mean = model.x_mean.value
         y_mean = model.y_mean.value
         x_stddev = model.x_stddev.value

astropy/units/quantity_helper/function_helpers.py

@@ -442,7 +442,7 @@ def select(condlist, choicelist, default=0):
 def piecewise(x, condlist, funclist, *args, **kw):
     from astropy.units import Quantity
 
-    # Copied implementation from numpy.lib.function_base.piecewise,
+    # Copied implementation from numpy.lib._function_base_impl.piecewise,
     # taking care of units of function outputs.
     n2 = len(funclist)
     # undocumented: single condition is promoted to a list of one condition

astropy/utils/masked/core.py

@@ -19,6 +19,7 @@
 
 import numpy as np
 
+from astropy.utils.compat import NUMPY_LT_2_0
 from astropy.utils.data_info import ParentDtypeInfo
 from astropy.utils.shapes import NDArrayShapeMethods
 
@@ -759,9 +760,17 @@ def __array_ufunc__(self, ufunc, method, *inputs, **kwargs):
             else:
                 # Parse signature with private numpy function. Note it
                 # cannot handle spaces in tuples, so remove those.
-                in_sig, out_sig = np.lib.function_base._parse_gufunc_signature(
-                    ufunc.signature.replace(" ", "")
-                )
+                if NUMPY_LT_2_0:
+                    in_sig, out_sig = np.lib.function_base._parse_gufunc_signature(
+                        ufunc.signature.replace(" ", "")
+                    )
+                else:
+                    (
+                        in_sig,
+                        out_sig,
+                    ) = np.lib._function_base_impl._parse_gufunc_signature(
+                        ufunc.signature.replace(" ", "")
+                    )
                 axis = kwargs.get("axis", -1)
                 keepdims = kwargs.get("keepdims", False)
                 in_masks = []

astropy/utils/masked/function_helpers.py

@@ -93,7 +93,7 @@
     np.isclose, np.allclose, np.flatnonzero, np.argwhere,
     # np.core.shape_base
     np.atleast_1d, np.atleast_2d, np.atleast_3d, np.stack, np.hstack, np.vstack,
-    # np.lib.function_base
+    # np.lib._function_base_impl
     np.average, np.diff, np.extract, np.meshgrid, np.trapz, np.gradient,
     # np.lib.index_tricks
     np.diag_indices_from, np.triu_indices_from, np.tril_indices_from,
@@ -108,7 +108,7 @@
     np.real_if_close, np.common_type,
     # np.lib.ufunclike
     np.fix, np.isneginf, np.isposinf,
-    # np.lib.function_base
+    # np.lib._function_base_impl
     np.angle, np.i0,
 }  # fmt: skip
 IGNORED_FUNCTIONS = {
@@ -121,7 +121,7 @@
 IGNORED_FUNCTIONS |= {
     np.pad, np.searchsorted, np.digitize,
     np.is_busday, np.busday_count, np.busday_offset,
-    # numpy.lib.function_base
+    # numpy.lib._function_base_impl
     np.cov, np.corrcoef, np.trim_zeros,
     # numpy.core.numeric
     np.correlate, np.convolve,
@@ -439,7 +439,7 @@ def msort(a):
 
 @dispatched_function
 def sort_complex(a):
-    # Just a copy of function_base.sort_complex, to avoid the asarray.
+    # Just a copy of np.lib._function_base_impl.sort_complex, to avoid the asarray.
     b = a.copy()
     b.sort()
     if not issubclass(b.dtype.type, np.complexfloating):  # pragma: no cover
@@ -595,11 +595,15 @@ def median(a, axis=None, out=None, **kwargs):
         )
         return (r.reshape(k) if keepdims else r) if out is None else out
 
-    else:
+    elif NUMPY_LT_2_0:
         return np.lib.function_base._ureduce(
             a, func=_masked_median, axis=axis, out=out, **kwargs
         )
 
+    return np.lib._function_base_impl._ureduce(
+        a, func=_masked_median, axis=axis, out=out, **kwargs
+    )
+
 
 def _masked_quantile_1d(a, q, **kwargs):
     """
@@ -608,7 +612,12 @@ def _masked_quantile_1d(a, q, **kwargs):
     """
     unmasked = a.unmasked[~a.mask]
     if unmasked.size:
-        result = np.lib.function_base._quantile_unchecked(unmasked, q, **kwargs)
+        if NUMPY_LT_2_0:
+            result = np.lib.function_base._quantile_unchecked(unmasked, q, **kwargs)
+        else:
+            result = np.lib._function_base_impl._quantile_unchecked(
+                unmasked, q, **kwargs
+            )
         return a.from_unmasked(result)
     else:
         return a.from_unmasked(np.zeros_like(a.unmasked, shape=q.shape), True)
@@ -641,7 +650,9 @@ def quantile(a, q, axis=None, out=None, **kwargs):
 
     a = Masked(a)
     q = np.asanyarray(q)
-    if not np.lib.function_base._quantile_is_valid(q):
+    if (NUMPY_LT_2_0 and not np.lib.function_base._quantile_is_valid(q)) or (
+        not NUMPY_LT_2_0 and not np.lib._function_base_impl._quantile_is_valid(q)
+    ):
         raise ValueError("Quantiles must be in the range [0, 1]")
 
     if NUMPY_LT_1_24:
@@ -650,11 +661,15 @@ def quantile(a, q, axis=None, out=None, **kwargs):
             a, func=_masked_quantile, q=q, axis=axis, out=out, **kwargs
         )
         return (r.reshape(q.shape + k) if keepdims else r) if out is None else out
-    else:
+    elif NUMPY_LT_2_0:
         return np.lib.function_base._ureduce(
             a, func=_masked_quantile, q=q, axis=axis, out=out, **kwargs
         )
 
+    return np.lib._function_base_impl._ureduce(
+        a, func=_masked_quantile, q=q, axis=axis, out=out, **kwargs
+    )
+
 
 @dispatched_function
 def percentile(a, q, *args, **kwargs):
@@ -761,7 +776,7 @@ def piecewise(x, condlist, funclist, *args, **kw):
     Any masks in ``condlist`` are ignored.
 
     """
-    # Copied implementation from numpy.lib.function_base.piecewise,
+    # Copied implementation from numpy.lib._function_base_impl.piecewise,
     # just to ensure output is Masked.
     n2 = len(funclist)
     # undocumented: single condition is promoted to a list of one condition