-
Notifications
You must be signed in to change notification settings - Fork 37
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
refactor: scaffolding to support custom context in extensions #816
base: main
Are you sure you want to change the base?
Conversation
NOTE: Since syrupy v4 migrated from instance methods to classmethods, this new context is not actual usable. This lays the groundwork for a switch back to instance methods though (if we continue along this path).
Thank you @noahnu for this PR! Would you mind letting me know how long this PR would take to merge? This feature was critical in some of our tests in TARDIS. If there is something I can do to help, please let me know!! |
@atharva-2001 I can't give an ETA. Could you describe what you're trying to do in your project (possibly with an example)? I may be able to recommend a workaround. |
I see. Here is some example code. Since most of my code deals with NumPy arrays and Pandas dataframes, I want to send in additional assertion options, for example, rtol, the assertion function etc. I don't want to create multiple fixtures by typing them out. Is it possible to at least create them programatically? Thanks for all the help! from typing import Any
import numpy as np
import pytest
from syrupy.data import SnapshotCollection
from syrupy.extensions.single_file import SingleFileSnapshotExtension
from syrupy.types import SerializableData
class NumpySnapshotExtenstion(SingleFileSnapshotExtension):
_file_extension = "dat"
def matches(self, *, serialized_data, snapshot_data, **kwargs):
print(kwargs, "kwargs inside matches")
try:
if (
# Allow multiple assertion methodds here, for example- assert_almost_equal
# allow relative and default tolerance
np.testing.assert_allclose(
np.array(snapshot_data), np.array(serialized_data), **kwargs
)
is not None
):
return False
else:
return True
except:
return False
def _read_snapshot_data_from_location(
self, *, snapshot_location: str, snapshot_name: str, session_id: str
):
# see https://github.com/tophat/syrupy/blob/f4bc8453466af2cfa75cdda1d50d67bc8c4396c3/src/syrupy/extensions/base.py#L139
try:
return np.loadtxt(snapshot_location).tolist()
except OSError:
return None
@classmethod
def _write_snapshot_collection(
cls, *, snapshot_collection: SnapshotCollection
) -> None:
# see https://github.com/tophat/syrupy/blob/f4bc8453466af2cfa75cdda1d50d67bc8c4396c3/src/syrupy/extensions/base.py#L161
filepath, data = (
snapshot_collection.location,
next(iter(snapshot_collection)).data,
)
np.savetxt(filepath, data)
def serialize(self, data: SerializableData, **kwargs: Any) -> str:
return data
@pytest.fixture
def snapshot_numpy(snapshot):
options = dict(matcher_options=dict(rtol=1, atol=0))
return snapshot.with_defaults(extension_class=NumpySnapshotExtenstion)
def test_np(snapshot_numpy):
x = [1e-5, 1e-3, 1e-1]
# ideally-
# from numpy.testing import assert_allclose
# assert snapshot_numpy(matcher = assert_allclose, rtol=1e6...)
assert snapshot_numpy == x |
@atharva-2001 Does something like this work until syrupy has built-in support? import pytest
class NumpySnapshotExtension(SingleFileSnapshotExtension):
_file_extension = "dat"
rtol = 0
atol = 0
def with_kwargs(**kwargs):
class MyCopy(NumpySnapshotExtension):
rtol = kwargs["rtol"]
atol = kwargs["atol"]
return MyCopy
@pytest.fixture
def snapshot(snapshot):
def factory(**kwargs):
_class = NumpySnapshotExtension.with_kwargs(**kwargs)
return snapshot.with_defaults(extension_class=_class)
return factory
def test_np(snapshot):
assert snapshot(rtol=1, atol=0) == [1e-5, 1e-3] (not tested) |
NOTE: Since syrupy v4 migrated from instance methods to classmethods, this new context is not actual usable. This lays the groundwork for a switch back to instance methods though (if we continue along this path).
Related to #814, this PR lays the groundwork to switch back to instance-based extensions (reverting an earlier decision to move to class methods for easier pytest-xdist compatibility).