pytorch/test/torch_np/test_scalars_0D_arrays.py

# Owner(s): ["module: dynamo"]

"""
Basic tests to assert and illustrate the  behavior around the decision to use 0D
arrays in place of array scalars.

Extensive tests of this sort of functionality is in numpy_tests/core/*scalar*

Also test the isscalar function (which is deliberately a bit more lax).
"""

from torch.testing._internal.common_utils import (
    instantiate_parametrized_tests,
    parametrize,
    run_tests,
    subtest,
    TEST_WITH_TORCHDYNAMO,
    TestCase,
    xfailIfTorchDynamo,
)


if TEST_WITH_TORCHDYNAMO:
    import numpy as np
    from numpy.testing import assert_equal
else:
    import torch._numpy as np
    from torch._numpy.testing import assert_equal


parametrize_value = parametrize(
    "value",
    [
        subtest(np.int64(42), name="int64"),
        subtest(np.array(42), name="array"),
        subtest(np.asarray(42), name="asarray"),
        subtest(np.asarray(np.int64(42)), name="asarray_int"),
    ],
)


@instantiate_parametrized_tests
class TestArrayScalars(TestCase):
    @parametrize_value
    def test_array_scalar_basic(self, value):
        assert value.ndim == 0
        assert value.shape == ()
        assert value.size == 1
        assert value.dtype == np.dtype("int64")

    @parametrize_value
    def test_conversion_to_int(self, value):
        py_scalar = int(value)
        assert py_scalar == 42
        assert isinstance(py_scalar, int)
        assert not isinstance(value, int)

    @parametrize_value
    def test_decay_to_py_scalar(self, value):
        # NumPy distinguishes array scalars and 0D arrays. For instance
        # `scalar * list` is equivalent to `int(scalar) * list`, but
        # `0D array * list` is equivalent to `0D array * np.asarray(list)`.
        # Our scalars follow 0D array behavior (because they are 0D arrays)
        lst = [1, 2, 3]

        product = value * lst
        assert isinstance(product, np.ndarray)
        assert product.shape == (3,)
        assert_equal(product, [42, 42 * 2, 42 * 3])

        # repeat with right-mulitply
        product = lst * value
        assert isinstance(product, np.ndarray)
        assert product.shape == (3,)
        assert_equal(product, [42, 42 * 2, 42 * 3])

    def test_scalar_comparisons(self):
        scalar = np.int64(42)
        arr = np.array(42)

        assert arr == scalar
        assert arr >= scalar
        assert arr <= scalar

        assert scalar == 42
        assert arr == 42


# @xfailIfTorchDynamo
@instantiate_parametrized_tests
class TestIsScalar(TestCase):
    #
    # np.isscalar(...) checks that its argument is a numeric object with exactly one element.
    #
    # This differs from NumPy which also requires that shape == ().
    #
    scalars = [
        subtest(42, "literal"),
        subtest(int(42.0), "int"),
        subtest(np.float32(42), "float32"),
        subtest(np.array(42), "array_0D", decorators=[xfailIfTorchDynamo]),
        subtest([42], "list", decorators=[xfailIfTorchDynamo]),
        subtest([[42]], "list-list", decorators=[xfailIfTorchDynamo]),
        subtest(np.array([42]), "array_1D", decorators=[xfailIfTorchDynamo]),
        subtest(np.array([[42]]), "array_2D", decorators=[xfailIfTorchDynamo]),
    ]

    import math

    not_scalars = [
        int,
        np.float32,
        subtest("s", decorators=[xfailIfTorchDynamo]),
        subtest("string", decorators=[xfailIfTorchDynamo]),
        (),
        [],
        math.sin,
        np,
        np.transpose,
        [1, 2],
        np.asarray([1, 2]),
        np.float32([1, 2]),
    ]

    @parametrize("value", scalars)
    def test_is_scalar(self, value):
        assert np.isscalar(value)

    @parametrize("value", not_scalars)
    def test_is_not_scalar(self, value):
        assert not np.isscalar(value)


if __name__ == "__main__":
    run_tests()