[MNT] Make aeon numpy compatible with both =>2.0 and <2.0

[TonyBagnall]

Reference Issues/PRs

closes #2218

test numpy2 compliance over simple fixes

STRAY np.Inf -> np.inf
Catch22: change list to array prior to call to nan_to_num

        c22_array = np.array(c22_list)
        if self.replace_nans:
            c22_array = np.nan_to_num(c22_array, False, 0, 0, 0)

performance_metrics.forecasting._functions

there is a difference in docstring tests: output of scalars appears as np.float64(x) not just x.

I have stripped out the expected output. Under numpy 2 it expects np.float64(0.0003), but that does not work with earlier numpy versions, where just 0.0003 is expected. the numbers are exactly the same

[aeon-actions-bot]

Thank you for contributing to aeon

I did not find any labels to add based on the title. Please add the [ENH], [MNT], [BUG], [DOC], [REF], [DEP] and/or [GOV] tags to your pull requests titles. For now you can add the labels manually.

The Checks tab will show the status of our automated tests. You can click on individual test runs in the tab or "Details" in the panel below to see more information if there is a failure.

If our pre-commit code quality check fails, any trivial fixes will automatically be pushed to your PR unless it is a draft.

Don't hesitate to ask questions on the aeon Slack channel if you have any.

PR CI actions

These checkboxes will add labels to enable/disable CI functionality for this PR. This may not take effect immediately, and a new commit may be required to run the new configuration.

• Run pre-commit checks for all files
• Run mypy typecheck tests
• Run all pytest tests and configurations
• Run all notebook example tests
• Run numba-disabled codecov tests
• Stop automatic pre-commit fixes (always disabled for drafts)
• Disable numba cache loading
• Push an empty commit to re-run CI checks

[TonyBagnall]

ok so isolated another weird difference arising in the test function test_stomp_squared_matrix_profile

this code

import numpy as np
expected = np.array([[18.00000,9.00000,6.00000,9.00000,18.00000,33.00000],
[19.00000,6.00000,5.00000,9.00000,6.00000,5.00000]])
id_bests = np.vstack(
    np.unravel_index(np.argsort(expected.ravel()), expected.shape)
).T

print(id_bests)

produces different output under numpy 1.26 and numpy 2.0. I assume due to tie breaking
1.26 output
[[1 2]
[1 5]
[0 2]
[1 1]
[1 4]
[0 1]
[0 3]
[1 3]
[0 0]
[0 4]
[1 0]
[0 5]]

2.0 output
[[1 5]
[1 2]
[0 2]
[1 1]
[1 4]
[1 3]
[0 3]
[0 1]
[0 4]
[0 0]
[1 0]
[0 5]]

[TonyBagnall]

Reminds me of the shapelet argsort issue under numba/no numba

see #622

[TonyBagnall]

simplified further

x= [18.00000,9.00000,6.00000,9.00000,18.00000,33.00000,19.00000,6.00000,5.00000,
    9.00000,6.00000,5.00000]
args = np.argsort(x)

gives different results in args under differnt numpy versions (and previously with and without numba #622). This is a very minor issue in one test, but we do use argsort quite a lot, might be time to implement our own version ...

[TonyBagnall]

ah they have added a "stable" argument! Problem solved :)
https://numpy.org/doc/2.0/reference/generated/numpy.argsort.html

[MatthewMiddlehurst]

Cant say I'm a fan of just removing all the expected outputs

[TonyBagnall]

Cant say I'm a fan of just removing all the expected outputs

if you dont you cannot be backward compatible as far as I can tell

[SebastianSchmidl]

The removed outputs are mostly in the performance metrics module. IMO, it has lower priority, so I don't care as much.

The problem is that the examples are checked using a string-comparison, right? Because np.float64(3.2) == 3.2 should return True:

• NumPy 1.x
  

• NumPy 2.x
  

[TonyBagnall]

The removed outputs are mostly in the performance metrics module. IMO, it has lower priority, so I don't care as much.

The problem is that the examples are checked using a string-comparison, right? Because np.float64(3.2) == 3.2 should return True.

agree, I think we skip ones we care about and just dont print of the performance metric module. I dont know how docstring runs the comparison, not really motivated to find out

changes made

[TonyBagnall]

aeon\similarity_search\matrix_profiles\tests\test_stomp.py:45 (test_stomp_squared_matrix_profile[3-int64])
dtype = 'int64', k = 3

    @pytest.mark.parametrize("dtype", DATATYPES)
    @pytest.mark.parametrize("k", K_VALUES)
    def test_stomp_squared_matrix_profile(dtype, k):
        """Test naive series search."""
        X = np.asarray(
            [[[1, 2, 3, 4, 5, 6, 7, 8]], [[1, 2, 4, 4, 5, 6, 5, 4]]], dtype=dtype
        )
    
        S = np.asarray([[3, 4, 5, 4, 3, 4, 5, 3, 2, 4, 5]], dtype=dtype)
        L = 3
        mask = np.ones((X.shape[0], X.shape[2] - L + 1), dtype=bool)
        distance = get_distance_function("squared")
        mp, ip = stomp_squared_matrix_profile(X, S, L, mask, k=k)
        for i in range(S.shape[-1] - L + 1):
            q = S[:, i : i + L]
    
            expected = np.array(
                [
                    [distance(q, X[j, :, _i : _i + L]) for _i in range(X.shape[-1] - L + 1)]
                    for j in range(X.shape[0])
                ]
            )
            id_bests = np.vstack(
                np.unravel_index(
                    np.argsort(expected.ravel(), kind="stable"), expected.shape
                )
            ).T
    
            for j in range(k):
                assert_almost_equal(mp[i][j], expected[id_bests[j, 0], id_bests[j, 1]])
>               assert_equal(ip[i][j], id_bests[j])

L          = 3
S          = array([[3, 4, 5, 4, 3, 4, 5, 3, 2, 4, 5]])
X          = array([[[1, 2, 3, 4, 5, 6, 7, 8]],

       [[1, 2, 4, 4, 5, 6, 5, 4]]])
distance   = CPUDispatcher(<function squared_distance at 0x000002D04F561820>)
dtype      = 'int64'
expected   = array([[20., 11.,  8., 11., 20., 35.],
       [21., 10.,  5., 11.,  8.,  3.]])
i          = 2
id_bests   = array([[1, 5],
       [1, 2],
       [0, 2],
       [1, 4],
       [1, 1],
       [0, 1],
       [0, 3],
       [1, 3],
       [0, 0],
       [0, 4],
       [1, 0],
       [0, 5]])
ip         = array([array([[0, 2],
              [1, 2],
              [1, 1]]), array([[1, 2],
                               [0, 2],
                               [1, 4]]), array([[1, 5],
                                                [1, 2],
                                                [1, 4]]), array([[1, 2],
                                                                 [0, 2],
                                                                 [0, 1]]),
       array([[0, 2],
              [1, 2],
              [1, 1]]), array([[1, 2],
                               [1, 5],
                               [1, 1]]), array([[1, 5],
                                                [1, 2],
                                                [0, 1]]), array([[0, 1],
                                                                 [1, 0],
                                                                 [0, 0]]),
       array([[0, 2],
              [1, 1],
              [0, 1]])], dtype=object)
j          = 2
k          = 3
mask       = array([[ True,  True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True,  True]])
mp         = array([array([0., 1., 2.]), array([2., 3., 3.]), array([3., 5., 8.]),
       array([2., 3., 4.]), array([0., 1., 2.]), array([5., 5., 6.]),
       array([ 9., 11., 13.]), array([2., 4., 5.]), array([1., 1., 2.])],
      dtype=object)
q          = array([[5, 4, 3]])

test_stomp.py:76: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
..\..\..\..\venv2\lib\site-packages\numpy\_utils\__init__.py:85: in wrapper
    return fun(*args, **kwargs)
        args       = (array([1, 4]), array([0, 2]), '', True)
        dep_version = '2.0.0'
        fun        = <function assert_array_equal at 0x000002D04D36CD30>
        kwargs     = {'strict': False}
        new_name   = 'desired'
        new_names  = ['actual', 'desired']
        old_name   = 'y'
        old_names  = ['x', 'y']
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

args = (<built-in function eq>, array([1, 4]), array([0, 2]))
kwds = {'err_msg': '', 'header': 'Arrays are not equal', 'strict': False, 'verbose': True}

    @wraps(func)
    def inner(*args, **kwds):
        with self._recreate_cm():
>           return func(*args, **kwds)
E           AssertionError: 
E           Arrays are not equal
E           
E           Mismatched elements: 2 / 2 (100%)
E           Max absolute difference among violations: 2
E           Max relative difference among violations: 1.
E            ACTUAL: array([1, 4])
E            DESIRED: array([0, 2])

args       = (<built-in function eq>, array([1, 4]), array([0, 2]))
func       = <function assert_array_compare at 0x000002D04D36CB80>
kwds       = {'err_msg': '',
 'header': 'Arrays are not equal',
 'strict': False,
 'verbose': True}
self       = <contextlib._GeneratorContextManager object at 0x000002D04D374220>

C:\Users\Tony\AppData\Local\Programs\Python\Python39\lib\contextlib.py:79: AssertionError


PASSED [ 73%]PASSED [ 80%]FAILED       [ 86%]
aeon\similarity_search\matrix_profiles\tests\test_stomp.py:45 (test_stomp_squared_matrix_profile[3-float64])
dtype = 'float64', k = 3

    @pytest.mark.parametrize("dtype", DATATYPES)
    @pytest.mark.parametrize("k", K_VALUES)
    def test_stomp_squared_matrix_profile(dtype, k):
        """Test naive series search."""
        X = np.asarray(
            [[[1, 2, 3, 4, 5, 6, 7, 8]], [[1, 2, 4, 4, 5, 6, 5, 4]]], dtype=dtype
        )
    
        S = np.asarray([[3, 4, 5, 4, 3, 4, 5, 3, 2, 4, 5]], dtype=dtype)
        L = 3
        mask = np.ones((X.shape[0], X.shape[2] - L + 1), dtype=bool)
        distance = get_distance_function("squared")
        mp, ip = stomp_squared_matrix_profile(X, S, L, mask, k=k)
        for i in range(S.shape[-1] - L + 1):
            q = S[:, i : i + L]
    
            expected = np.array(
                [
                    [distance(q, X[j, :, _i : _i + L]) for _i in range(X.shape[-1] - L + 1)]
                    for j in range(X.shape[0])
                ]
            )
            id_bests = np.vstack(
                np.unravel_index(
                    np.argsort(expected.ravel(), kind="stable"), expected.shape
                )
            ).T
    
            for j in range(k):
                assert_almost_equal(mp[i][j], expected[id_bests[j, 0], id_bests[j, 1]])
>               assert_equal(ip[i][j], id_bests[j])

L          = 3
S          = array([[3., 4., 5., 4., 3., 4., 5., 3., 2., 4., 5.]])
X          = array([[[1., 2., 3., 4., 5., 6., 7., 8.]],

       [[1., 2., 4., 4., 5., 6., 5., 4.]]])
distance   = CPUDispatcher(<function squared_distance at 0x000002D04F561820>)
dtype      = 'float64'
expected   = array([[20., 11.,  8., 11., 20., 35.],
       [21., 10.,  5., 11.,  8.,  3.]])
i          = 2
id_bests   = array([[1, 5],
       [1, 2],
       [0, 2],
       [1, 4],
       [1, 1],
       [0, 1],
       [0, 3],
       [1, 3],
       [0, 0],
       [0, 4],
       [1, 0],
       [0, 5]])
ip         = array([array([[0, 2],
              [1, 2],
              [1, 1]]), array([[1, 2],
                               [0, 2],
                               [1, 4]]), array([[1, 5],
                                                [1, 2],
                                                [1, 4]]), array([[1, 2],
                                                                 [0, 2],
                                                                 [0, 1]]),
       array([[0, 2],
              [1, 2],
              [1, 1]]), array([[1, 2],
                               [1, 5],
                               [1, 1]]), array([[1, 5],
                                                [1, 2],
                                                [0, 1]]), array([[0, 1],
                                                                 [1, 0],
                                                                 [0, 0]]),
       array([[0, 2],
              [1, 1],
              [0, 1]])], dtype=object)
j          = 2
k          = 3
mask       = array([[ True,  True,  True,  True,  True,  True],
       [ True,  True,  True,  True,  True,  True]])
mp         = array([array([0., 1., 2.]), array([2., 3., 3.]), array([3., 5., 8.]),
       array([2., 3., 4.]), array([0., 1., 2.]), array([5., 5., 6.]),
       array([ 9., 11., 13.]), array([2., 4., 5.]), array([1., 1., 2.])],
      dtype=object)
q          = array([[5., 4., 3.]])

test_stomp.py:76: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
..\..\..\..\venv2\lib\site-packages\numpy\_utils\__init__.py:85: in wrapper
    return fun(*args, **kwargs)
        args       = (array([1, 4]), array([0, 2]), '', True)
        dep_version = '2.0.0'
        fun        = <function assert_array_equal at 0x000002D04D36CD30>
        kwargs     = {'strict': False}
        new_name   = 'desired'
        new_names  = ['actual', 'desired']
        old_name   = 'y'
        old_names  = ['x', 'y']
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

args = (<built-in function eq>, array([1, 4]), array([0, 2]))
kwds = {'err_msg': '', 'header': 'Arrays are not equal', 'strict': False, 'verbose': True}

    @wraps(func)
    def inner(*args, **kwds):
        with self._recreate_cm():
>           return func(*args, **kwds)
E           AssertionError: 
E           Arrays are not equal
E           
E           Mismatched elements: 2 / 2 (100%)
E           Max absolute difference among violations: 2
E           Max relative difference among violations: 1.
E            ACTUAL: array([1, 4])
E            DESIRED: array([0, 2])

args       = (<built-in function eq>, array([1, 4]), array([0, 2]))
func       = <function assert_array_compare at 0x000002D04D36CB80>
kwds       = {'err_msg': '',
 'header': 'Arrays are not equal',
 'strict': False,
 'verbose': True}
self       = <contextlib._GeneratorContextManager object at 0x000002D04D374220>

C:\Users\Tony\AppData\Local\Programs\Python\Python39\lib\contextlib.py:79: AssertionError