monkeypatch.setenv does not work consistently using cli

[lachhebo]

• [ x] a detailed description of the bug or problem you are having

monkeypatch does not work reliably as it should. Actually, my test is running well well launched from pytest gui but not when run with cli.

• [ x] output of pip list from the virtual environment you are using:

Package Version

---

api-wafermark 0.1.0
appdirs 1.4.4
attrs 21.2.0
bcrypt 3.2.0
certifi 2020.12.5
cffi 1.14.5
cfgv 3.2.0
chardet 4.0.0
click 7.1.2
coverage 5.5
cycler 0.10.0
distlib 0.3.1
ecdsa 0.14.1
fastapi 0.63.0
filelock 3.0.12
flake8 3.9.2
freezegun 1.1.0
glob2 0.7
h11 0.12.0
identify 2.2.4
idna 2.10
iniconfig 1.1.1
kiwisolver 1.3.1
Mako 1.1.4
MarkupSafe 2.0.0
matplotlib 3.4.2
mccabe 0.6.1
mypy 0.812
mypy-extensions 0.4.3
nodeenv 1.6.0
numpy 1.20.3
packaging 20.9
parse 1.19.0
parse-type 0.5.2
passlib 1.7.4
Pillow 8.2.0
pip 21.0.1
pluggy 0.13.1
pre-commit 2.12.1
py 1.10.0
pyasn1 0.4.8
pycodestyle 2.7.0
pycparser 2.20
pydantic 1.8.2
pyflakes 2.3.1
pyparsing 2.4.7
pytest 6.2.4
pytest-bdd 4.0.2
pytest-cov 2.11.1
python-dateutil 2.8.1
python-jose 3.2.0
python-multipart 0.0.5
PyYAML 5.4.1
requests 2.25.1
rsa 4.7.2
setuptools 56.0.0
six 1.16.0
starlette 0.13.6
toml 0.10.2
typed-ast 1.4.3
typing-extensions 3.10.0.0
urllib3 1.26.4
uvicorn 0.13.4
virtualenv 20.4.6
wheel 0.36.2

• pytest and operating system versions

macos catalina with pytest 6.2.4

• minimal example if possible

This test:

from datetime import timedelta, datetime
from unittest.mock import patch

from freezegun import freeze_time

TESTED_MODULE = 'api_wafermark.security.access_token'


@freeze_time("2012-01-14")
@patch(f'{TESTED_MODULE}.jwt.encode')
def test_create_access_token__expires_in_defined_time(mock_jwt_encode, monkeypatch):
    # given

    monkeypatch.setenv('JWT_SECRET_KEY', 'SECRET_K')
    monkeypatch.setenv('ALGORITHM', 'ALGO')
    monkeypatch.setenv('DB_URL', 'postgresql://user:pass@localhost:5432/wafermark')

    from api_wafermark.security.access_token import create_access_token

    data_ = {'secret': 'isma is the best'}
    treated_data = {'secret': 'isma',
                    'exp': datetime.utcnow() + timedelta(minutes=15)
                    }

    # when
    create_access_token(data_, 15)

    # then
    mock_jwt_encode.assert_called_with(treated_data, 'SECRET_K', algorithm='ALGO')

For this file:

import os
from datetime import datetime, timedelta
from typing import Optional, Dict

from jose import jwt

SECRET_KEY = os.getenv('JWT_SECRET_KEY')
ALGORITHM = os.getenv('ALGORITHM', 'HS256')
ACCESS_TOKEN_EXPIRE_MINUTES = os.getenv('ACCESS_TOKEN_EXPIRE_MINUTES', 60)


def create_access_token(data: Dict, expires_delta: Optional[int] = ACCESS_TOKEN_EXPIRE_MINUTES):
    to_encode = data.copy()
    expire = datetime.utcnow() + timedelta(minutes=expires_delta)
    to_encode.update({"exp": expire})
    encoded_jwt = jwt.encode(to_encode, SECRET_KEY, algorithm=ALGORITHM)
    return encoded_jwt

i have this error:

>           raise AssertionError(_error_message()) from cause
E           AssertionError: expected call not found.
E           Expected: encode({'secret': 'isma', 'exp': FakeDatetime(2012, 1, 14, 0, 15)}, 'SECRET_K', algorithm='ALGO')
E           Actual: encode({'secret': 'isma', 'exp': FakeDatetime(2012, 1, 14, 0, 15)}, None, algorithm='HS256')

[RonnyPfannschmidt]

this is a user error

monkeypatching the environment after you pulled the values will not change the pulled values

the suggestion would be to either monkeypatch the module directly and/or to do on demand loading of the env vars

[lachhebo]

I know, that is why i import the module after using monkeypatch, most of the time it worked but not this time which is kind of strange.

(it is working when i run the test using pycharm but not when i use the cli)

[The-Compiler]

I'm converting this to a discussion as this definitely isn't an issue with pytest, like @RonnyPfannschmidt said.

Imports in Python are cached, so likely something else in your tests is already importing the module earlier. Consider e.g. adding 1/0 to the module, to provoke a ZeroDevisionError and find out where/how that happens.

[nicoddemus]

@The-Compiler's suggestion to find when the module is imported is on point, but I would like to suggest to avoid using globals here completely:

def get_token_params() -> tuple[str, str, int]:
    secret = os.environ['JWT_SECRET_KEY']
    algorithm = os.environ.get('ALGORITHM', 'HS256')
    expire = int(os.environ.get('ACCESS_TOKEN_EXPIRE_MINUTES', 60))
    return secret, algorithm, expire


def create_access_token(data: Dict, expires_delta: Optional[int] = None):
	secret, algorithm, expire = get_token_params()
	if expires_delta is None:
		expires_delta = expire
    to_encode = data.copy()
    expire = datetime.utcnow() + timedelta(minutes=expires_delta)
    to_encode.update({"exp": expire})
    encoded_jwt = jwt.encode(to_encode, secret, algorithm=algorithm)
    return encoded_jwt

This will avoid the problem altogether, even preventing this problem from reappearing in the future in case some imports move around.