Issue upgrading `pydata/sparse` to v2025.12

[prady0t]

Upgrading to v2025.12 in pydata/sparse, gives a weird error for test_floor_divide[__ifloordiv__(x1, x2)] test. See the issue: pydata/sparse#943

_____________________________________ test_floor_divide[__ifloordiv__(x1, x2)] ______________________________________
[gw3] darwin -- Python 3.13.14 /Users/pradyotranjan/sparse-1/.pixi/envs/xp-tests/bin/python3.13
  + Exception Group Traceback (most recent call last):
  |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 1376, in test_floor_divide
  |     @given(data=st.data())
  |                    ~~~~^^
  |   File "/Users/pradyotranjan/sparse-1/.pixi/envs/xp-tests/lib/python3.13/site-packages/hypothesis/core.py", line 2274, in wrapped_test
  |     raise the_error_hypothesis_found
  | ExceptionGroup: Hypothesis found 2 distinct failures. (2 sub-exceptions)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 1393, in test_floor_divide
    |     binary_param_assert_against_refimpl(ctx, left, right, res, "//", operator.floordiv)
    |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 680, in binary_param_assert_against_refimpl
    |     binary_assert_against_refimpl(
    |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
    |         func_name=ctx.func_name,
    |         ^^^^^^^^^^^^^^^^^^^^^^^^
    |     ...<10 lines>...
    |         strict_check=strict_check,
    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^
    |     )
    |     ^
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 349, in binary_assert_against_refimpl
    |     assert isclose(scalar_o, expected, M), msg
    | AssertionError: x1=3.0, but should be roughly (x1 // x2)=1.0 [__ifloordiv__()]
    |   x1=3.0, x2=2.0
    | assert False
    |  +  where False = isclose(3.0, 1.0, 1.7976931348623157e+308)
    | 
    | ========== FAILING CODE SNIPPET:
    | __ifloordiv__(<COO: shape=(), dtype=float64, nnz=1, fill_value=0.0>, <COO: shape=(), dtype=float32, nnz=1, fill_value=0.0>)
    | ====================
    | 
    | Falsifying example: test_floor_divide(
    |     # The test always failed when commented parts were varied together.
    |     ctx=BinaryParamContext(<__ifloordiv__(x1, x2)>),
    |     data=data(...),
    | )
    | Draw 1 (x1): _f(<...>, False)
    | Draw 2 (x2): _f(
    |     <...>,
    |     False,  # or any other generated value
    | )
    +---------------- 2 ----------------
    | Traceback (most recent call last):
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 1393, in test_floor_divide
    |     binary_param_assert_against_refimpl(ctx, left, right, res, "//", operator.floordiv)
    |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 680, in binary_param_assert_against_refimpl
    |     binary_assert_against_refimpl(
    |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
    |         func_name=ctx.func_name,
    |         ^^^^^^^^^^^^^^^^^^^^^^^^
    |     ...<10 lines>...
    |         strict_check=strict_check,
    |         ^^^^^^^^^^^^^^^^^^^^^^^^^^
    |     )
    |     ^
    |   File "/Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py", line 351, in binary_assert_against_refimpl
    |     assert scalar_o == expected, (
    |     ...<2 lines>...
    |     )
    | AssertionError: x1=-1, but should be (x1 // x2)=1 [__ifloordiv__()]
    |   x1=-1, x2=-1
    | assert -1 == 1
    | 
    | ========== FAILING CODE SNIPPET:
    | __ifloordiv__(<COO: shape=(), dtype=int8, nnz=1, fill_value=0>, <COO: shape=(), dtype=int8, nnz=1, fill_value=0>)
    | ====================
    | 
    | Falsifying example: test_floor_divide(
    |     # The test always failed when commented parts were varied together.
    |     ctx=BinaryParamContext(<__ifloordiv__(x1, x2)>),
    |     data=data(...),
    | )
    | Draw 1 (x1): _f(
    |     <...>,  # or any other generated value
    |     False,
    | )
    | Draw 2 (x2): _f(
    |     <...>,
    |     False,  # or any other generated value
    | )
    | Explanation:
    |     These lines were always and only run by failing examples:
    |         /Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py:352
    |         /Users/pradyotranjan/sparse-1/array-api-tests/array_api_tests/test_operators_and_elementwise_functions.py:354

array_api_tests/test_operators_and_elementwise_functions.py::test_floor_divide[__ifloordiv__(x1, x2)] - ExceptionGroup: Hypothesis found 2 distinct failures. (2 sub-exceptions)

[ev-br]

With hypothesis, it's indeed sometimes tricky to find out what's going on. To make it easier, the test suite emits the exact code constructs which fail, see the FAILING CODE SNIPPET below:

    | AssertionError: x1=3.0, but should be roughly (x1 // x2)=1.0 [__ifloordiv__()]
    |   x1=3.0, x2=2.0
    | assert False
    |  +  where False = isclose(3.0, 1.0, 1.7976931348623157e+308)
    | 
    | ========== FAILING CODE SNIPPET:
    | __ifloordiv__(<COO: shape=(), dtype=float64, nnz=1, fill_value=0.0>, <COO: shape=(), dtype=float32, nnz=1, fill_value=0.0>)
    | ====================

The snippet is generated by this line https://github.com/data-apis/array-api-tests/blob/master/array_api_tests/test_operators_and_elementwise_functions.py#L1387 --- it uses repr but apparently repr is not very helpful for pydata/sparse objects.

So I'd suggest to locally modify the output to actually print out the values of the array arguments. This way you'll be able to see if the floor division is actually correct. This will show where the problem is, in the test suite or in pydata/sparse.