inline-snapshot-numpy

NumPy integration for inline-snapshot.

Installation

pip install inline-snapshot-numpy

Usage

The package provides snapshot-aware versions of NumPy's array assertions:

from inline_snapshot_numpy import assert_array_equal
from numpy import array


def test_array():
    actual = array([[1, 2], [3, 4]])
    assert_array_equal(actual)

Create the missing snapshot with:

pytest --inline-snapshot=create

The test will become:

def test_array():
    actual = array([[1, 2], [3, 4]])
    assert_array_equal(actual, desired=array([[1, 2], [3, 4]]))

The following helpers are available: assert_array_equal, assert_allclose, assert_array_almost_equal, assert_array_almost_equal_nulp, and assert_array_max_ulp.

The expected array is managed through snapshot_arg(), so no explicit snapshot() call or import is needed. You can still pass an explicit snapshot(...) for compatibility.

Alternatively, call setup() from conftest.py to patch the corresponding functions in numpy.testing:

from inline_snapshot_numpy import setup

setup()

Arrays can also be stored in external .npy snapshot files:

from inline_snapshot import external
from inline_snapshot_numpy import assert_array_equal
from numpy import array


def test_array():
    assert_array_equal(array([1, 2, 3]), external())

Create the external snapshot with pytest --inline-snapshot=create. NumPy object arrays are intentionally rejected because loading them would require enabling pickle. Snapshot reports render the array shape, dtype, and values as text, and array changes are shown as a unified text diff instead of a binary hexdump.

License

Distributed under the terms of the MIT license.