ENH: allow python scalars in the 2nd argument of searchsorted

The upcoming Array API revision 2025.12 will allow the second argument of
searchsorted to be a python scalar, cf
data-apis/array-api#982

Author: ev-br

cupy/_sorting/search.py

@@ -430,7 +430,10 @@ def _searchsorted(a, v, side, sorter, assume_increasing):
         raise NotImplementedError('Only int or ndarray are supported for a')
 
     if not isinstance(v, cupy.ndarray):
-        raise NotImplementedError('Only int or ndarray are supported for v')
+        if not isinstance(v, int | float | complex):
+            raise NotImplementedError(
+                'Only python scalars or ndarrays are supported for v')
+        v = cupy.asarray(v, dtype=a.dtype)
 
     if a.ndim > 1:
         raise ValueError('object too deep for desired array')

tests/cupy_tests/sorting_tests/test_search.py

@@ -699,7 +699,8 @@ def test_searchsorted(self, xp, dtype):
         x = testing.shaped_arange(self.shape, xp, dtype)
         bins = xp.array(self.bins)
         y = xp.searchsorted(bins, x, side=self.side)
-        return y,
+        y1 = xp.searchsorted(bins, 2, side=self.side)   # python scalar for `v`
+        return y, y1
 
     @testing.for_all_dtypes(no_bool=True)
     @testing.numpy_cupy_array_equal()