unit-testing

[hexed cloak]

Or remind someone to unskip the test when they add the feature

[primal dagger]

Hi, has anyone come across that while running specifi tests with unit tests, the test passed, but while running all of the tests, some of them were failing?

[hexed cloak]

Yep

[potent quest]

yes

[hexed cloak]

Try using pytest-randomly can help discover test interdependence and test isolation problems

Eg one of your other tests didn't clean up after itself

[primal dagger]

What if I have to use unittests library?

Because it is a part of code competition

and code checker uses unittests

so i wonder if there is anything I can do to fix that

[hexed cloak]

Is the purpose of the challenge to find out why the tests fail in this way?

[primal dagger]

Nope

I have some test cases provided

And running them seperately all of them are passing

but if I run all of them at once

SOme of them are failing

[hexed cloak]

You could try and find the first git commit where it started falling in this way?

[potent quest]

hm, are you allowed to use pytest?

[primal dagger]

Nope, the code checker uses unittest

[potent quest]

you don't have to rewrite the tests, but pytest can give you additional information

ah

[primal dagger]

It was not passing before, so probably I will need to rewrite the code

[potent quest]

!rule 8 ? @hexed cloak?

[bitter wadi]

Rules

8. Do not help with ongoing exams. When helping with homework, help people learn how to do the assignment without doing it for them.

[potent quest]

competition == ongoing exam maybe?

[primal dagger]

I mean, i do not need to write my code

[hexed cloak]

It's still pretty vague so I think ok

Could be anything though

[potent quest]

oh

oftentimes these checkers are horrible

[hexed cloak]

I've seen a bunch of failures like this in tests that didn't rollback the database, or a mock.patch

Or they did a manual patch

[potent quest]

you're reusing a variable somewhere and one of the tests is checking that state

[hexed cloak]

Or a file not in a tmp_path

Or they relied on random.Random seed

[primal dagger]

Ahhh

okay

got it

[hexed cloak]

Had a good one where they left a thread running

[primal dagger]

It was because oop and because the object was not teared down, it was overwitten in the other test

That is my assumption

Thank you guys

[hexed cloak]

?

Best to make a fresh object each time

Rather than try to restore anything

[primal dagger]

Yeah, but probably because the name of the obejct stayed the same in each test case, it could be overwitten, as this Solution class did not have dunder init method

Kinda my mistake

[hexed cloak]

?

[harsh storm]

Hi.. needed some help with my project structure

├───lib
│   ├───discord_bot
│   │   ├─── __main__.py
│   │   ├─── __init__.py 
│   │   ├─── headers.py 
│   │   └───cogs
│   │         └───helpers
│   │              └───Common
│   │                  ├─── __init__.py
│   │                  ├─── constants.py
│   │                  └─── some_api.py
│   ├───resources
│   └───tests
│         ├─── __init__.py
│         └─── test_some_api.py
└───scripts

With above folder structure, I have headers.py at the discord bot folder and inside some_api.py, I have this import

from headers import InstaHeaders as headers

How would I re-write that import so that it doesn't fail in my test file?

[hexed cloak]

from discord_bot.headers import InstaHeaders as headers

you also need cogs/__init__.py and cogs/helpers/__init__.py

[fiery arrow]

I have a project with decent unit tests, but pretty much no testing on the edge, the endpoints and stuff. Any tips on where to start?

[maiden pawn]

Selenium is the most developed web driver controller with the support for the most amount of browsers including edge

if by edge you mean Microsoft Edge

if by edge u mean integration testing... than in my opinion it is almost no different than writing unit tests

same pytest library (+ framework specific testing library / or requests library if nothing else ) is more than enough to do that

we write in same way to check those endpoints, in some cases we mock them usually when they are third party dependency
I prefer to be not mocking endpoints which belong to my backend services.

[maiden pawn]

Ah, and the most important part.

when testing ednpoints stuff between your own microservices

that's where CI/CD stuff shines the best (like Gitlab CI)

when u set with some Infrastructure Provisioning like Terraform automatically to buy/set everything in cloud provider, and to install your services with some Configuration Management like Ansible + Docker for containrization

when everything is deployed automatically with CI/CD to run unit tests, deploy to staging environment and then to apply integration tests... that's the best / easiest / the least consuming way to handle it

[fiery arrow]

no, by "the edge" I don't mean Microsoft Edge

so the problem with having no integration tests is that the tests don't actually test if the system works

my question is: where do I get started, writing-tests-wise, if I want to write "real" tests, that don't mock/stub/fake anything out?

[hexed cloak]

depends what the project does

and how it runs in production

just write tests that call the system as it would be in production without any stubs and see where it crashes because you don't have a service in your test environment

[pearl cliff]

there are also tools for doing that kind of thing, e.g. Cypress seems to be a popular choice

but it's not python

[harsh storm]

This causes issues... I get this error on startup:

    from headers import InstaHeaders as headers
ModuleNotFoundError: No module named 'discord_bot'
.
.
.
discord.ext.commands.errors.ExtensionFailed: Extension 'cogs.utilities' raised an error: ModuleNotFoundError: No module named 'discord_bot'```

[hexed cloak]

You need to run it with python -m discord_bot

And you have to do from discord_bot.headers import InstaHeaders as headers

[harsh storm]

Sorry, had added the above import itself. I had edited the error logs to cover up some names.

[hexed cloak]

Ok well don't do that

[harsh storm]

python -m discord_bot

How does above work?

[hexed cloak]

It runs your discord_bot.__main__ file and sets up your Python sys.path correctly

[harsh storm]

Ohh okay, I'll try that and let you know

[gaunt zodiac]

So it sets CWD to same directory where discord_bot is ?

Or am i misunderstanding smthing ?

[hexed cloak]

it sets sys.path for you so you can do "import discord_bot" or "from . import whatever" and it all works

[gaunt zodiac]

Oh, alright

[idle grotto]

you guys have any recommended resources for learning unit testing? And should I ALWAYS unit test my projects, even if it's a small script?

[daring shore]

I'm looking for feedback on some tests I wrote, please would someone let me know if they're okay, if they need improvement, etc?

import requests
import unittest

from datetime import datetime

url = "http://127.0.0.1:8000"

def to_iso(datetime_object):
    return str(datetime_object).replace(" ", "T")


class CreateUser(unittest.TestCase):
    @classmethod
    def setUpClass(cls):
        cls.unique_string = "u6kat2ldn1"
        # fmt: off
        cls.test_user = {
            'user_id': -1, 'first_name': 'TestUser', 'last_name': 'TestSurname', 'email': '[e-mail]',
            'biography': cls.unique_string, 'date_of_birth': f"{to_iso(datetime(2000, 1, 1))}",
            'creation_date': f"{to_iso(datetime(2021, 1, 1))}",
            'pronouns': ["he", "him", "his", "his", "himself"], 'phone_number': '023 9282 0660'.replace(' ', ''),
        }
        # Make a copy of the dict above
        cls.expected_user = dict((key, value) for key, value in cls.test_user.items())  
        del cls.expected_user["user_id"]
        # fmt: on

    def setUp(self):
        print("Setting up CreateUser")
        r = requests.delete(url + f"/delete_user_by_bio/{self.unique_string}")
        self.assertEqual(r.status_code, 200)

    def test_create_user(self):
        print("Testing CreateUser")
        request = requests.post(url + "/create_user", json=self.test_user)
        self.assertEqual(request.status_code, 200)
        self.assertTrue("users" in request.json())
        user = request.json()["users"][0]
        self.assertTrue("user_id" in user)
        del user["user_id"]
        self.assertEqual(user, self.expected_user)

    def tearDown(self):
        print("Tearing down CreateUser")
        r = requests.delete(url + f"/delete_user_by_bio/{self.unique_string}")
        self.assertEqual(r.status_code, 200)


if __name__ == "__main__":
    unittest.main()

Please ping me :)
And uhh, when creating a user, I set the user_id to -1 for API reasons, but want to ignore it because it uses AUTO_INCREMENT in the database, so it's returned id is unknown

[fiery arrow]

If the 8000 port is occupied, the test will break

[daring shore]

Hmm, good point! How did I fix that? I probably need to make everything asynchronous

[idle grotto]

Thanks Darkwind

[lapis gale]

it's often beneficial to write the tests before the actual code

test driven development

[idle grotto]

how do you decide what tests to write exactly? thats where im stumped

like how to actually start

[maiden pawn]

the book answers that

and there are book recommendations to continue readings, about TDD included

The book the first point is to explain when and how to write tests

in order for you to give understanding, which tests are helping with minimum maintance effort and maximum result

[lapis gale]

guess you might find other resources than a 40 euro book

the topic touches software-design a lot

[maiden pawn]

one 40 euro book, teaches materials usually learned for half of a year in online courses/universities and e.t.c.
comparing time and money spent on book with other alternatives... it is shortest and most time efficient way to learn stuff

[lapis gale]

and since the name is 'unit-testing' it seems that it doesnt cover 'higher' tests

[maiden pawn]

it covers

[lapis gale]

as a start, whenever you find a bug, write a minimal test example that will fail due to this bug and fix it until the test runs smoothly

[idle grotto]

hmmm. that logic sounds backwards

but ill do some research and get some clarity

thank you both!

[maiden pawn]

then more forward logic:
When u need to write new functionality/feature, write it first in a test.
Then copy result(import function from same place) to main program ;b

Or even better.
When u write new feature... write test first describing what that feature is without implementing it yet.
Then start thinking how to implement it while keeping in mind how to test it the best/the easiest way
Implement it and have it program and testable in a test to pass the test.

[waxen holly]

if I've got a couple files I need to test against, where should I store them?

[hexed cloak]

what do you mean by against?

[waxen holly]

with open("test_collection.db", "r") as test, open("default_collection.db", "r") as default):
    self.assertEqual(test.read(), default.read())```

test_collection being the resulting file from running the test

[hexed cloak]

I store files with my tests like

numpee_as_np
├── core
│   ├── foo.py
│   ├── __init__.py
│   └── test
│       ├── example.csv
│       ├── __init__.py
│       └── test_foo.py
├── __init__.py
└── __main__.py

then I do:

import importlib.resources

files = importlib.resources.files(__package__)

def test_foo():
    example = (files / "example.csv").read_text(encoding="utf8")

[waxen holly]

why importlib?

[hexed cloak]

to import resources

from the package

[waxen holly]

I got that from the code lol, why use that over open?

[hexed cloak]

because open looks in the cwd

and importlib.resources.files(__package__) looks in the package

and works pretty much no matter how your package is installed

eg it could be a zip file or an egg or a whl or whatever

[waxen holly]

gotcha

[hexed cloak]

also if you're just doing:

with open(...) as f:
    v = f.read()

it's better to use .read_text(...)

eg v = (files / "some_file").read_text(encoding="utf8")

[waxen holly]

ok cool good to know

[dense bough]

And if you are reading from a resource, there is even importlib.resources.read_text(__package__, "example.csv")!

Hmm... I remember reading that sub-directories are not supported yet. Wonder if that has changed yet.

[hexed cloak]

That's deprecated

They basically deprecated everything in there when they added the files function

[dense bough]

Really? I don't see it in the docs.

[hexed cloak]

Check the 3.11 version

[dense bough]

Ah..

Looks like I have some grep-ing to do.

This is going to be a pain to sieve through...

Thanks for the heads-up about the deprecation!

[hexed cloak]

sieve through? there's literally one function you can call now .files(__package__)

[dense bough]

I mean, to sieve through code using the deprecated function.

And updating them.

[hexed cloak]

ah I see

yeah basically

honestly "open_text" and "read_text" has an ok interface

@dense bough also one to watch out for .files().open_text encoding default to the filesystem encoding and not utf8

[frigid basalt]

That's the case for open() as well.. super scary to think I've basically been omitting it when I should've used it

[dusty sigil]

can you help?

    def name(self):
        return self.__name

    @name.setter
    def name(self, value):
        if not value.isalpha():
            raise ValueError("Team Name can contain only letters!")
        self.__name = value```

how do I unit test this?

[hexed cloak]

Show your full code and use ```python

[dusty sigil]

class Team:
    def __init__(self, name: str):
        self.name = name
        self.members: Dict[str, int] = {}

    @property
    def name(self):
        return self.__name

    @name.setter
    def name(self, value):
        if not value.isalpha():
            raise ValueError("Team Name can contain only letters!")
        self.__name = value

    def add_member(self, **name_age):
        added_members_by_name = []
        for name, age in name_age.items():
            if name not in self.members:
                self.members[name] = age
                added_members_by_name.append(name)
        return f"Successfully added: {', '.join(added_members_by_name)}"

    def remove_member(self, name: str):
        if name in self.members:
            del self.members[name]
            return f"Member {name} removed"
        else:
            return f"Member with name {name} does not exist"

    def __gt__(self, other):  # "other" is another instance of class Team!
        if len(self.members) > len(other.members):
            return True
        return False

    def __len__(self):
        return len(self.members)

    def __add__(self, other):  # "other" is another instance of class Team!
        new_team_name = f"{self.name}{other.name}"
        new_team = Team(new_team_name)
        new_team.add_member(**self.members)
        new_team.add_member(**other.members)
        return new_team

    def __str__(self):
        result = [f"Team name: {self.name}"]
        members = list(sorted(self.members.items(), key=lambda x: (-x[1], x[0])))
        result.extend([f"Member: {x[0]} - {x[1]}-years old" for x in members])
        return "\n".join(result)

this is the code

i need unit test for this

can you help?

[hexed cloak]

Why didn't you use ```python ?

[dusty sigil]

i used it

[hexed cloak]

def foo() -> None:
    pass

That's what happens if you use it

Screenshot the message before you send it?

[dusty sigil]

ok

[hexed cloak]

You need a newline between python and from

[dusty sigil]

from typing import Dict


class Team:
    def __init__(self, name: str):
        self.name = name
        self.members: Dict[str, int] = {}

    @property
    def name(self):
        return self.__name

    @name.setter
    def name(self, value):
        if not value.isalpha():
            raise ValueError("Team Name can contain only letters!")
        self.__name = value

    def add_member(self, **name_age):
        added_members_by_name = []
        for name, age in name_age.items():
            if name not in self.members:
                self.members[name] = age
                added_members_by_name.append(name)
        return f"Successfully added: {', '.join(added_members_by_name)}"

    def remove_member(self, name: str):
        if name in self.members:
            del self.members[name]
            return f"Member {name} removed"
        else:
            return f"Member with name {name} does not exist"

    def __gt__(self, other):  # "other" is another instance of class Team!
        if len(self.members) > len(other.members):
            return True
        return False

    def __len__(self):
        return len(self.members)

    def __add__(self, other):  # "other" is another instance of class Team!
        new_team_name = f"{self.name}{other.name}"
        new_team = Team(new_team_name)
        new_team.add_member(**self.members)
        new_team.add_member(**other.members)
        return new_team

    def __str__(self):
        result = [f"Team name: {self.name}"]
        members = list(sorted(self.members.items(), key=lambda x: (-x[1], x[0])))
        result.extend([f"Member: {x[0]} - {x[1]}-years old" for x in members])
        return "\n".join(result)```

Done

[hexed cloak]

Yay

[dusty sigil]

can you help now>

?

[hexed cloak]

You forgot type annotations on most of your methods?

[dusty sigil]

from project.team import Team

import unittest

class TestStringMethods(unittest.TestCase):
    _TESTNAME = "TestName"
    _TESTNEWNAME=  "alabala"
    _TESTDICT = {}

    def setUp(self) -> None:
        self.test_shop = Team(self._TESTNAME)

    def test_init(self):
        self.assertEqual(self._TESTNAME,self. test_shop.name)
        self.assertDictEqual(self._TESTDICT, self.test_shop.members)

    def test_remove_member(self):
        result = self.test_shop.remove_member(self._TESTNEWNAME)
        expected = f"Member with name {self._TESTNEWNAME} does not exist"
        self.assertEqual(result, expected)




if __name__ == '__main__':
    unittest.main()

this is my unit test

[hexed cloak]

What tests do you have so far?

[dusty sigil]

so far

[hexed cloak]

Yeah

Currently what tests do you have?

Oh I see your post

Honestly there's no need to bother with a setUp for this sort of system under test

Just make a new Team in the body of any test

Also all of those constants are pointless - the constant is longer than the value

I'd probably write

def test_name_invalid():
    t = Team("ham")
    with pytest.raises(ValueError, match=r"Team Name can only contain letters!"):
        t.name = "2"
    assert t.name == "ham"

And another for a valid name

[dusty sigil]

how can I change all "_TESTNAME" to "test"

in one click

and not one by one

[hexed cloak]

Depends on the editor

Ctrl+h does it for me

[dusty sigil]

im on pycharm

[hexed cloak]

Use the pycharm "help -> find action" tool

[dusty sigil]

didn't help

😫

[hexed cloak]

Screenshot?

[hexed cloak]

Did you try Rename all occurrences?

[dense bough]

The -m version allows current directory to be searched for test modules.

[hexed cloak]

Usually you don't want to use python -m pytest

[dense bough]

Are there any recommended ways to unit test custom pytest fixtures?

Specifically, I want to test the fixture clean-up.

Nevermind. I can probably make it work with pytester.

[rare wolf]

hey, can we parameterize a fixture list

@pytest.fixture()
def blah():
    return [1, 2, 3, 4]

@pytest.mark.parametrize('a', blah)
def test_a(a):
    assert a == 1

i know it's possible without fixture

[frigid basalt]

Just name the parameter blah instead of a and that code will work

[rare wolf]

play_with_test.py:None (play_with_test.py)
Fixture "blah" called directly. Fixtures are not meant to be called directly,
but are created automatically when test functions request them as parameters.
See https://docs.pytest.org/en/stable/fixture.html for more information about fixtures, and
https://docs.pytest.org/en/stable/deprecations.html#calling-fixtures-directly about how to update your code.

@pytest.fixture()
def blah():
    return [1, 2, 3, 4]

@pytest.mark.parametrize('blah', blah())
def test_a(blah):
    assert blah == 1

i got an error

[frigid basalt]

Remove the parametrize

[rare wolf]

what do you mean

@frigid basalt what do you mean?

[frigid basalt]

@pytest.fixture()
def blah():
    return [1, 2, 3, 4]

def test_a(blah):
    assert blah == 1

Pytest will call the fixture

[rare wolf]

@frigid basalt , it's work but i want to test each element seperatly

@frigid basalt i have mentioned that i want it as parametrize

[frigid basalt]

Do you need blah to be a fixture? You could just have it be a normal function you call.

[rare wolf]

i know that too

but in my case i have a dynamic fixture

@frigid basalt i know that too

[frigid basalt]

I am a bit confused, I don't know how to help.

[rare wolf]

@frigid basalt I'm confused too, I don't know how to solve it

[plush vessel]

😋

[proper wind]

#black-formatter

[burnt whale]

Can I do not pass @mock.patch('...', return_value=...) mock as an argument to an invoked function?
Why? I don't need it. Moreover - among other 10 mocks it is hard to maintain the correct order.
Is _ the only option?

[hexed cloak]

Use the context manager interface

You can put all 10 patch calls in the same with statement

But that's a lot of patches, maybe write your tests without patch instead?

[burnt whale]

Isn't it much readable.

Yea, I of course know that's so many patched, but it's unavoidable

[hexed cloak]

It's much readable yes

[burnt whale]

Ooo, with some mocks (that we care about) as @ and the others in with statement - yeaa, better

[hexed cloak]

The context manager lets you directly associate the patch and the value without having to count up args backwards

[shut wave]

using hypothesis:

import hypothesis.strategies as strats


@given(strats.text(), strats.integers())
def test_base_charset(text, integers):                                
    """The characters in the base charspace must come from the charset
    properties."""                     
    sp = Charspace(
        blocklist=text,
        extras=text,                       
        uppers=text,   
        lowers=text,               
        digits=str(integers),
    )                                                                 
    everything = sp.extras + sp.uppers + sp.lowers  # + sp.digits
                
    assert all(char in everything for char in sp.base_charset)        

Is my use of strategies.text() giving me different text for different properties, or is it going to be the same on each pass? I assume the latter. Do I need to do strats.text(), strats.text(), strats.text(), strats.text(), like this or is there a better way?

[silk turret]

You use property not right, you dont need name mangling here. Use it like this. Its common pattern, when u create private attribute in constructor, and open calculated public property.
' py
class Team:
def init(self, name: str):
self._name = name
self.members: Dict[str, int] = {}

@property
def name(self):
    return self._name

'

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @calm oak until <t:1639643438:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[dusty sigil]

my question is how to do a unit test on this

[dusty sigil]

i don't want you to fix the code

[deep plover]

Can anyone help me out with an issue I'm having with unit tests, mocking/patching and assert_called_with: https://stackoverflow.com/questions/70381238/pytest-mocker-patch-is-returning-noncallabalemagicmock

[hexed cloak]

anyone know how to mock.path("path.to.some_async_fn", side_effect=some_other_async_fn) ?

[hexed cloak]

I sort of expected it to await the some_other_async_fn

!e

import pytest
from unittest import mock

class Foo:
    async def bar(self):
        pass

@pytest.mark.anyio
async def test_foo():
    async def fake_bar():
        return 0

    with mock.patch.object(Foo, "bar", side_effect=fake_bar):
        assert await Foo().bar() == 1

pytest.main()

[bitter wadi]

@hexed cloak :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 15, in <module>
003 |   File "/usr/local/lib/python3.10/asyncio/runners.py", line 44, in run
004 |     return loop.run_until_complete(main)
005 |   File "/usr/local/lib/python3.10/asyncio/base_events.py", line 641, in run_until_complete
006 |     return future.result()
007 |   File "<string>", line 13, in test_foo
008 | AssertionError

[hexed cloak]

!e

import pytest
from unittest import mock

class Foo:
    async def bar(self):
        pass

@pytest.mark.anyio
async def test_foo():
    async def fake_bar():
        return 0

    with mock.patch.object(Foo, "bar", side_effect=fake_bar):
        assert await Foo().bar() == 1

pytest.main()

ah dang

[frigid basalt]

!e ```python
import pytest
from unittest import mock

class Foo:
async def bar(self):
pass

@pytest.mark.anyio
async def test_foo():
async def fake_bar():
return 1 # This was set to 0?

with mock.patch.object(Foo, "bar", side_effect=fake_bar):
    assert await Foo().bar() == 1

pytest.main()

[bitter wadi]

@frigid basalt :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 1, in <module>
003 | ModuleNotFoundError: No module named 'pytest'

[frigid basalt]

Oh?

[frigid basalt]

How come this got far enough

[hexed cloak]

Oh I edited it xD

But too late for the rerurn

[frigid basalt]

Oh lol

[hexed cloak]

Oh lol I got the wrong number in there

But when I run it locally I get a Coroutine back from await Foo().bar()

!e

import anyio
from unittest import mock

class Foo:
    async def bar(self):
        pass

async def test_foo():
    async def fake_bar():
        return 1

    with mock.patch.object(Foo, "bar", side_effect=fake_bar):
        assert await Foo().bar() == 1

anyio.run(test_foo)

[bitter wadi]

@hexed cloak :warning: Your eval job has completed with return code 0.

[No output]

[hexed cloak]

Huh

It does work

That's very odd maybe I'm just using an old version of mock

[frigid basalt]

That would make sense

[frigid basalt]

Yeah I saw that was like..

[hexed cloak]

smh

[frigid basalt]

https://c.tenor.com/-RHLoZc1J-QAAAAM/nick-young-question-mark.gif

[neon moss]

Not sure if this belong in here, but here we go. I'm trying to use cibuildwheels to build wheels, and there are some tests that I can get to run but fail because the files it relies on lives in a different directory. Anyone come up with the "best" way of handling this?

[kind meadow]

Which file(s) does the test rely on? Why are they in a different directory than expected? How is it related to cibuildwheels? Are you using this https://cibuildwheel.readthedocs.io/en/stable/options/#test-command ?

[neon moss]

Yeah, it's a cibuildwheels thing. There's files in /examples/samples/ that are part of the test suite. cibuildwheels uses a new venv to test wheels in and doesn't use/have access to the project folder. For linux I managed to get it to copy them across by using this;
CIBW_BEFORE_TEST_LINUX: mkdir -p ~/examples/samples && cp -R {project}/examples/samples/* ~/examples/samples/ however I stalled out with windows :/

[kind meadow]

How about making your tests read the path from an environment variable. Then you can just set the environment variable to a path relative to {project} and cibuildwheel should be able to replace that with the actual path, right?

You could also give it a default value for when you want to run tests locally.

os.getenv(TEST_DATA_PATH, "examples/samples")

[neon moss]

Hmm that's a great idea

[ebon pewter]

Any suggestions on the best way to test data pipelines accurately?

[pearl cliff]

what kinds of data pipelines?

if the pipeline is "plain" python code, then you can test the python code as usual

testing "data processing" code can be difficult, because it can be difficult to come up with invariant properties to test

for ETL systems e.g. using airflow, i'm not sure what the industry standard is... but at minimum you can run some sample data through the full pipeline and make sure it doesn't return any errors, the number of rows is correct, the column names are all correct, check for data types, etc.

it seems like silly obvious stuff, but you never know when you introduce a typo that breaks something

[kind meadow]

I have some code that needs to be tested in specific environments. I'm not used to testing at this "higher level". How would I provision these environments for automated tests? For example, I need environments with certain cgroup version, swap memory on and off, and the swap memory cgroup controller on and off.

Furthermore, let's say I somehow already had all these environments ready to go. How should I approach writing the actual tests? The expected behaviour I need to assert will differ depending on the environment. Should I write a separate test for each environment that only runs if in that environment?

[potent quest]

smh i wondered what this was related to at first read lol

[pearl cliff]

maybe you need to set up a bunch of virtual machine images. and you can set a special environment variable or create a specific file that indicates the current test environment

e.g. MYAPP_TEST_ENV=foo123 or echo foo123 > /etc/myapp/test-env

[kind meadow]

Yeah I was thinking something along those lines.

Does anyone have practical experience with this? Anyone have tests that rely on specific environments that can't feasibly be mocked? Would love to hear how others have set it up.

I suppose a popular form of this is OS-specific code. CI mostly supports this use case out of the box though. Setting up runners with specific Linux kernel parameters... not so much.

[strong briar]

Hi, how does one go about testing your application's HTTP requests? Referring to actually testing the whole stack, including the network call, and I'm comparing with Java where there is WireMock, a HTTP server that you set up during testing to respond to your http requests locally

I'm yet to find a similar library for python, they all seem to 'mock' the http request and don't actually make the API call. is this the norm with python?

This is more end-to-end / integration testing territory but I didn't find a more appropriate channel

Only httpretty seems to be close to that but I'm not sure it does what I need to either

Or if I should even be bothering at this level and should just care that the endpoint is called without my application actually doing the http request at runtime

The other one is responses

[empty dust]

Do you mean something like https://pytest-httpserver.readthedocs.io/en/latest/tutorial.html#tutorial

[strong briar]

exactly what I was looking for. However, I have 2 problems

1. I'm using unittest instead of pytest, is it a good practise to use both unittest and pytest as I can't immediately find a unittest alternative?
2. I'm using an api wrapper library which talks to real apis ( test environment & live ) and i might not be able to set the base url to localhost

but this is what I was looking for regardless, thank you

which is basically talked about in the docs you sent:

Some code changes required in the original source code. The code should accept the server endpoint (host and port) as a parameter or by some means of configuration. This endpoint will be set to localhost during the test running. If it is not possible, you need to tweak name resolution.

I'll see if i can make that happen although it's not possibly out of the box in the library im using

[empty dust]

1. No issue with that, you might have to make some changes to get it to work (not sure if there's an example on the wiki)
2. That's mentioned in the how-to https://pytest-httpserver.readthedocs.io/en/latest/howto.html#customizing-host-and-port

[strong briar]

Thank you so much!

I guess I need to improve my googling skills, should have been able to find this

[deep plover]

Can anyone help me with mocking a class with property and setters than access a "private" attribute: I've outlined the problem here: https://stackoverflow.com/questions/70464598/pytest-mocking-property-and-setter

[cinder valve]

i am stuck

anyone available?

[strong briar]

Hi, similar to my question above, I'm now looking on testing my WebSocket client. It simply connects to a WS server and subscribes to a topic. I want to write a test for this that would ideally start a fake ws server and send a message to the topic that my app would consume

Is there a python library that achieves somethings along these lines? Ideally not mocking but actually making use of the network TCP layer

[strong briar]

Trying aiohttp's websocket & test support, hope it might do it. If anyone has feedback please provide it 🙂

https://github.com/aio-libs/aiohttp/blob/master/tests/test_web_websocket_functional.py

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @regal folio until <t:1640431334:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[proper wind]

Hello, is it possible to unit testing web scraper project? Do I need to mock the html response first?

[empty dust]

For unit testing a web scraper your best bet will probably be to download a webpage locally and open it with BeautifulSoup (see https://stackoverflow.com/a/21577649)

You can test all parts of your code with that - mocking http responses will depend on which request library you are using (e.g. for httpx there is https://pypi.org/project/pytest-httpx/)

[proper wind]

Hi, thanks for the answer.

Right now I'm using Bottle micro framework, I'm still confused how to do the unit test without rewriting the scraping algorithm, do I need to create another separate function and calls it inside bottle default function so that I can use it later in test?

Thanks again.

I'm already see the unit-testing approach in Bottle documentation: https://bottlepy.org/docs/dev/recipes.html#:~:text=them in production.-,UNIT-TESTING BOTTLE APPLICATIONS,-¶, but the problem is the test will be flaky because this test doesn't mock the response of the scraping request.

[empty dust]

Hard to say without being able to see your code, but in general to test code you have to design it to be testable. That might mean you need to shift things around

[proper wind]

Solved using responses library to mock requests, thanks.

[woeful wasp]

Hi everyone

[bitter wadi]

Hey @spiral geyser!

It looks like you tried to attach a Python file - please use a code-pasting service such as https://paste.pythondiscord.com

[spiral geyser]

https://github.com/BlueJayCool7788/Math-Math.py

[kind meadow]

I'm having trouble testing this function. I want to test that the arguments given via CLI are actually used. Is this just a case where I need to mock the function and assert the args its called with? I've been trying to avoid that.

def main():
    args = parse_args()
    result = NsJail().python3(args.code, nsjail_args=args.nsjail_args, py_args=args.py_args)
    print(result.stdout)

    if result.returncode != 0:
        sys.exit(result.returncode)

[potent quest]

you could mock NsJail, sys.exit instead, I suppose

[kind meadow]

How is that different than what I said

[potent quest]

i thought you were trying to mock main() 🤦‍♀️

that's all on me tho

[maiden pawn]

so this function can be invoked via CLI?
they don't mock anything, call it in your test externally via os.system or smth like that
and check its response in stdout 😉 (better to use subprocess.run with capturing stdout I guess then)

lets make it as much black box testing as possible

mocks should be avoided when it is possible, since they make tests fragile

[kind meadow]

The problem is that I can't find a good way of ensuring the arguments are passed in a black box way

Testing their behaviour is not straight forward

Nsjail args are passed to a subprocess, which in turn spawns a Python subprocess which uses the py args.

[fiery arrow]

so you want to make an integration test with nsjail?

I guess you could test individual things, like --binding a temporary folder to a known location and then reading form that known location in the Python code.

you'll have to test the "real thing" eventually, right? you can't mock forever

[kind meadow]

I do test the real thing, but through the python3 directly rather than the CLI entrypoint (which is mainly there for debugging/development purposes)

@fiery arrow

So it doesn't necessarily have to be an integration test, but it would be nice to not tightly couple tests to code with mocks.

I can test the return code easily. I can test that it prints stdout easily. The arguments are just the problem, particularly the Python arguments. For NsJail args at least I can set a time limit and execute code that sleeps longer than the limit.

[potent quest]

wdym python args?

[kind meadow]

https://docs.python.org/3/using/cmdline.html

[fiery arrow]

You can just use sys.argv in the Python code, can't you?

or maybe I'm misunderstanding

[kind meadow]

It's interpreter args

[fiery arrow]

oh, it's the arguments to python itself

[potent quest]

well, python 3.10 has an attr containing them and there's another flag for that too in earlier versions

[fiery arrow]

@kind meadow try -O and check that an assert False doesn't work

[potent quest]

!d sys.flags

[bitter wadi]

sys.flags

sys.flags```
The [named tuple](https://docs.python.org/3/glossary.html#term-named-tuple) *flags* exposes the status of command line flags. The attributes are read only...

[fiery arrow]

or that lol yes, TIL

[kind meadow]

And ideally, I could create a test case that passes both nsjail and py args, otherwise my test won't tell me if they work together (even though I do have unit tests for parse_args already)

[potent quest]

!d sys.orig_argv

[bitter wadi]

sys.orig\_argv

sys.orig_argv```
The list of the original command line arguments passed to the Python executable.

See also [`sys.argv`](https://docs.python.org/3/library/sys.html#sys.argv "sys.argv").

New in version 3.10.

[kind meadow]

Thanks, that might work

[potent quest]

I've never read through all of sys's docs on a whim 🤡

[frosty timber]

Hey, I need to start testing this application https://github.com/HicaroD/tChat

To be very honest, I don't really the mindset of a tester, I just don't know how to start. I was trying to figure out where my application could break and the only thing that comes in my mind is network problems, in case of the internet doesn't work, the program just crash.

So, any tips about how to test it or even tips about testing in general

[slender kernel]

Talking about pytest, if I have two fixtures in a test function:

1. What is the evaluation order? Is it left to right?
2. What if both of them have a yield? Will both of the fixture stop until test is run to teardown?

[potent quest]

1: that's indeterminate with just that info, as fixtures can depend on each other.
2: yes

[slow stump]

trying to do unit tests in django keeps returning "0 tests found"

[maiden pawn]

What is your testing framework

[frosty timber]

https://docs.djangoproject.com/en/4.0/topics/testing/overview/

I think it's a good start

It teaches you how to write and run tests in Django

[vocal phoenix]

hey guyzz

is it fine?

i have actually created a weight converter..

weight = int(input('Weight: '))
unit = input('(L)bs or (K)g: ')
if unit.upper() == "L":
converted = weight * 0.45
print(f'You are {converted} kilos')
else:
converted = weight / 0.45
print(f'You are {converted} pounds')

using pycharm

[open galleon]

Hello ! I'm wondering if there is any Selenium user here.. I'm following a tutorial on the net which uses the method "find_element_by_xpath" but PyCharm is angry bcs the method seems deprecated.. What would be the elegant way to find a button using his text without find_element_by_xpath ?

[hexed cloak]

https://stackoverflow.com/a/69875984/833093

[open galleon]

Thanks a lot ! 🙂

[fierce flare]

Selenium/Python/pytest users could benefit a lot from https://github.com/seleniumbase/SeleniumBase , which is a framework for simplifying Selenium with Python.

[bronze vortex]

from selenium.webdriver.common.by import By

find_element(By.XPATH, '//xpath')```

[strong briar]

Does anyone know of a way to test an application that connects to a WebSocket server? Namely, being able to send messages from the WebSocket server running during testing to the application under test, not just the WS server replying to messages sent by the application.

[frigid basalt]

Is this for a Discord API library?

[strong briar]

Hi, no. It's for a cryptocurrency project, listening for coin price data

[open galleon]

Someone know why send_keys is not sending the whole key sometimes ? I did not have this issue while using Chrome, know when I'm sending some keys, a lot of the text inside is not sent. It also looks like for some unknown reasons Selenium is doing a Ctrl + F and writing my text inside it lol..

[jaunty hearth]

wdym?

[proper wind]

@open galleon yes.

[frigid basalt]

How do you guys deal with optional tests?

I have some tests in another package that I want to import over - but this package is optional - but I am not sure how to handle it missing. Reporting one warning for say 10 missing tests doesn't seem like reflects what's missing in a sense?

[pearl cliff]

@frigid basalt can you use unittest.skipIf?

try:
    import optional_dep
except ImportError:
    _optional_dep_available = False
else:
    _optional_dep_available = True

@unittest.skipUnless(_optional_dep_available, 'Skipping because optional_dep is unavailable.')
class TestOptionalDepStuff(unittest.TestCase):
    ...

[frigid basalt]

Oh I didn't realize that "skip" was what I was looking for, awesome!

[candid gate]

Is TDD something that someone fairly inexperienced should try and incorporate? Totally willing to, but it will definitely slow me down (for better or worse. Might actually be for better since I can think about what I want to write before writing it)

[pearl cliff]

it might be worth trying it a few times. it might help clarify your thinking and designs. personally i don't do it, because i find that i don't usually have the APIs fleshed out until after i've already started writing the code, and if i write the tests too early in the process i end up having to change them later anyway

but if you write some small utility functions and you know what the interfaces will be, then sure, go ahead and write the tests first

but it's worth at least trying it. some people really enjoy programming that way

one argument in favor of TDD is that you would spend that time debugging anyway

[potent quest]

yes but now you also debug the tests too

🥲

[pearl cliff]

true

test suites can become their own software projects, depending on what you are building

[potent quest]

. @frigid basalt definitely was doing this today

[frigid basalt]

😔

[maiden pawn]

Sure. Actually TDD can help the most to unexperienced people.
Because it FORCES you to think architecture of your code in a clean and testable way 😉

Actually I don't even always follow TDD, but I still keep following the main point of it, thinking in how i am going to test my code first before writing it

[ember maple]

testable and clean are not necessarily the same thing tho - a very interesting example of this is the stark test philosophy and refactoring philosophy differences of mercurial and bzr

[maiden pawn]

testable enforses minimal clean code quality nevertherless, I agree that not fully and not equal to clean, but still quite considerable leap forward
At least for me quality of tested code is raised at least several times always (even 10 times feels right)

[ember maple]

using tests as driver tends to be a big help, but the choosen detail levle is a thing of its own

[pearl cliff]

what's the difference in these philosophies? i have never used either, although i am familiar with comparisons between git and hg

i know approximately nothing about bzr

[ember maple]

mercurial is tested using integration tests, the outer facade being the cli with ruthless refactorings of the internals, bzr is tested using quite a number of unittests, it has massive amounts of internal interfaces

[pearl cliff]

ahh interesting

yeah two very different styles. great illustration

[ember maple]

a key part however is - if you do mercurial style testing, not doing ruthless refactoring will lead to massive spagetty

[pearl cliff]

the linux kernel also famously follows the former, right? although it also famously doesn't have tests...

[ember maple]

the linux kernel is rather strange with the testing strategy, i cant comment too much on that

im just using linux cause its the least painful to me, not cause i think its a good os ^^

i once naively tried to make a "good" os, it was a good lesson in learning not to overestimate oneself

[pearl cliff]

heh

i use linux because of the operating systems built on top of it, not because i have any particular affinity for any particular kernel

i am not enough of a power user to care significantly about the differences between the bsd kernel and the linux kernel... i just know how to use linux-based systems and most software runs on them. good enough for me

[ember maple]

its hard to be fond of anything thats build on top of the posix permission/file model ^^

[pearl cliff]

i don't have a reference point for anything better 🙂

[ember maple]

but we are getting slightly OT ^^

in beos there was a little attempt ^^ but basically file systems are a pain that jsut store blobs with minimal metadata (which is why so many file formats are also metadata containers ^^)

[pearl cliff]

ahh yeah

i am aware that beos did some fun stuff like that

and that before unix you had pretty complicated filesystems, and switching to 'blobs of bytes' was apparently a big improvement at the time

[ember maple]

thats for sure, problem is that schemas and schema validation + evolution is very expensive ^^

[rare hill]

Suggestions for the best way to mock a couple levels down into an object?

        cs = self.cnx.cursor()
        result = []

        try:
            blah = cs.execute(query)
            result = list(blah)
        finally:
            cs.close()

I want to test this method, meaning I need to have the two methods it calls (cnx.cursor() and .execute() on the object returned by cursor()) mocked, right? What's the best way to do that?

[pearl cliff]

i think i'd probably do something like this

obj = TheClassWithTheMethodYouWantToTest()
with patch.object(obj, 'cnx') as mock_cnx:
    mock_cursor = MagicMock(return_value=[ ... ])
    mock_cnx.cursor = MagicMock(return_value=mock_cursor)
    obj.execute_query()
    mock_cursor.execute.assert_called_once_with(query)
    mock_cursor.close.assert_called_once_with()

[rare hill]

What I've currently got is

            expected_result = ['1','2','3']

            execute_mock = Mock()
            execute_mock.return_value = expected_result

            cursor_mock = Mock()
            cursor_mock.execute = execute_mock
            cursor_mock.close = Mock()

            connection_mock = Mock()
            connection_mock.cursor = cursor_mock
                        
            snowflake_connector_mock.return_value = connection_mock

            client = SnowflakeClient(self.settings)
    
            result = client.execute_query('SELECT * FROM DATA')

            self.assertEqual(result, expected_result)

But that seems to return a mock object from the execute() method rather than my expected_result list

[pearl cliff]

where the [ ... ] is some fake result data

ah, the classic

you have to patch where it's used

obj = TheClassWithTheMethodYouWantToTest()
with patch.object(obj, 'cnx') as mock_cnx:
    mock_cursor = MagicMock()
    if query_should_fail:
        mock_cursor.execute = MagicMock(side_effect=QueryFailureException())
    else:
        mock_cursor.execute = MagicMock(return_value=fake_data)
    mock_cnx.cursor = MagicMock(return_value=mock_cursor)

    obj.execute_query()
    mock_cursor.execute.assert_called_once_with(query)
    mock_cursor.close.assert_called_once_with()

[rare hill]

Ohhh you mean where it's consumed in the actual class I'm testing

I have read about that but still don't have it understood well enough

[pearl cliff]

yes

but in this case it sounds like you want to patch one specific object anyway instead of the whole class

[rare hill]

Yea I need to mock methods on an instance

[pearl cliff]

but also this seems like you're patching deep into SnowflakeClient which is 3rd party code

and you usually don't want to patch into 3rd party code if you can avoid it

[rare hill]

Yeah...honestly I'm only writing these tests because SonarCloud requires 80% coverage. IMO this doesn't even really need a test! But I have to add something. So I figured I can mock the Snowflake connection stuff, have it return some specific data, then just verify that my function invokes the snowflake connection to get that data

[pearl cliff]

yeah it's not bad to test "small" methods imo

but just be careful about what you're mocking, whether it depends on internals of other packages or not

[rare hill]

I think my larger issue is not understanding why it doesn't get the return value back in this instance

[pearl cliff]

@rare hill where does snowflake.connector.connect actually get used?

did you write SnowflakeClient or is that 3rd party code?

i assume that's the class you are testing?

you need to do the patch in the module where SnowflakeClient is defined

[frigid basalt]

I am in pretty big pain and looking for how to parametrize my tests or somehow reduce code. I have a suite of ratelimiting tests. I want to test a bunch of combinations of:

• Same/different ratelimit bucket (a key in a dictionary)
• Same/different endpoint (an instance of a class)
• Same/different major parameter (value into the instantiation of the former class)

I also want to make some tests without the bucket, because that may be missing.

A typical test looks like this: https://paste.pythondiscord.com/ofenumujor.sql

I've been thinking about what/how I can parametrize but I am not sure if anyone has any ideas on how to structure my tests?

I want to test essentially all possible combinations of same/different XYZ. Which becomes:

# 12 tests:
# - S stands for Same, D for different
# Bucket, Endpoint, Major parameter

# SB, SE, SM
# DB, SE, SM - SB, DE, SM - SB, SE, DM    # 1 different (3 tests separated by the '-')
# DB, DE, SM - SB, DE, DM - DB, SE, DM    # 2 different
# DB, DE, DM

# SE, SM
# DE, SM - SE, DM    # 1 different
# DE, DM

The sample test I showed tests SB, DE, SM

Should I just look to replace the changing parts with variables or any ideas of how I can test it differently?

[fiery arrow]

ngl, every time I see "Permission denied" I'm afraid I destoy some bearing wall with my head

worst error of all time

[midnight bramble]

If I am testing a class method and I need to iterate through many different instances (with different attributes) and iterate over many method parameters, what is the easiest most pythonic way to do that?

...or resources for doing that? Most unit testing resources seem focused on testing parameters.

[pearl cliff]

if you are using unittest, it sounds like you can just write a loop and use subTest. if you are using pytest, it sounds like you want parametrize

[frigid basalt]

Update on this, I have somewhat succeeded! Now, I realize that there's even more combinations I want to test (I want very thorough coverage and considering I know there's a finite number of combinations I would really like to write good tests that test this)..

For example, different major parameter value but also different major parameter at all (only possible with different endpoints, maybe not worth it then?) would be another set of combinations I haven't thought about. Am I testing too much and being over the top or is there maybe any idea how I can generate these combinations during runtime when the tests are loaded so it doesn't look so messy?

Here's how it looks right now, this tests the latter comments that doesn't consider buckets: https://paste.pythondiscord.com/qisuzepayi.py

[kind meadow]

What are some nice ways to deal with long test data for pytest's parametrize decorator? I don't find multi-line decorators to be particularly readable. Should the data be put in another Python file and then imported? What other options have you seen that you've liked or done yourself?

[maiden pawn]

Moving to another file sounds about right.
As long as tests and file dependencies lie outside of folder with main code

Tests should not be in the way of main code

Or as an option, you could try moving this big test in a separated file 😉

I have two approaches to keep tests

1. Django way, when tests.py file at the same place where the package of the code

2. test files in separate folder outside of the main code

in second option i try to name then precisely with sub categories like
tests_category_thing.py

makes them organized more or less
but I guess all those file organizations depend on amount of present code/tests

TLDR: moving just data to external file sounds right, because it is usually comfortable to have tests of the same category easily scrolled together 🤔
It should be more comfortable than having big test in a separate file

[glass pivot]

pretty sure he wasn't asking about whether tests should be in a separate file, but rather whether the data for the parametrized decorator from pytest library should be in one @maiden pawn

[maiden pawn]

i voiced alternatives to it and compared with first suggested approach

[kind meadow]

Thanks

[dense ore]

Hi, i'm new to unit testing and I wanted to test my Quart app with pytest, but can't figure out how to do so. Tried following some documentations but got nowhere. Basically at first I tried making a directory containing all my test files and tying out a simple test on a function written in the service directory.
|---->App------|
|------->routes
|------->services------>file which contains the function to be tested
|------->tests------->file containing test cases
but when I tried to run tests using pytest I faced some import errors, then I got a solution which looked kind of hacky but worked out, basically by adding init.py in every directory the import issue was fixed. But them came along another challenge, adding support for globals.
Using pytest with quart is kind of confusing and I'm stuck. Don't know how to write tests for things in such a complex service, where there is a heavy use of globals, db queries and complex logics.
Any suggestions would mean a ton 🤞

[maiden pawn]

reading carefully
https://realpython.com/absolute-vs-relative-python-imports/

and doing something like...
going explicitely to folder from which you are launching your app
and launching test from the same folder

and writing imports in your tests in absolute starting from that folder or in relative way

considering your case, relative imports should work fine too 🤔

import will look like...

tests.py

from ..services.file1 import abc

def test1():
    assert abc==123

file1.py

abc=123

__init__.py files are just empty files

runs fine

(venv) naa@naa-MS-7C89:~/repos/experiments$ pytest
=============================================================================== test session starts ===============================================================================
platform linux -- Python 3.8.10, pytest-6.2.5, py-1.11.0, pluggy-1.0.0
rootdir: /home/naa/repos/experiments, configfile: pytest.ini
collected 1 item                                                                                                                                                                  

app/tests/tests.py .    

i provided relative way

absolute will look like

tests.py

from app.services.file1 import abc

def test1():
    assert abc==123

check carefully where i was launching the tests
naa@naa-MS-7C89:~/repos/experiments$ pytest
it is a parent folder to app folder

[frigid basalt]

How would I write a marker/fixture to run a test twice?

For example, let's pretend I want to test my comparison operator implementation:

@pytest.mark.parametrize(
    ['a', 'b', 'returns'],
    [
        (1, 2, False)
    ]
)
def test_compare(a, b, returns):
    assert (a == b) is returns

I want to run the test both with test_compare(1, 2, False) and test_compare(2, 1, False). Any tips?

[maiden pawn]

import pytest

@pytest.mark.parametrize(
    "a, b, returns",
    [(1, 2, False), (2, 1, False)],
)
def test1(a, b, returns):
    assert (a == b) is returns

[kind meadow]

I believe you just have to duplicate the tuple and flip the values

[frigid basalt]

How come I can use @pytest.mark.anyio and it will run my test twice (albeit with different event loops)?

There seems to be some mechanism to run a test multiple times

[kind meadow]

Do you want to make your own decorator to do that? Is that what you're getting at?

[frigid basalt]

I have a substantial amount of parametrize cases, I would prefer to not repeat it.

[frigid basalt]

Yes, that's what I was looking to do.

I don't know where to start with it though, because most fixtures just provide some value that's used - they don't run the test multiple times like parametrize

[kind meadow]

I'm not sure how to execute the test multiple times

If anyio does it you can look at its source, or look at parametrize's source

Looks like it's related to Metafunc.parametrize

[frigid basalt]

After looking at it for a while I found this: https://github.com/agronholm/anyio/blob/master/src/anyio/pytest_plugin.py

[kind meadow]

I guess you can just decorate it multiple times with parametrize

and make a wrapper decorator to hide that stuff

[frigid basalt]

But it's unclear to me where it's called multiple times

[frigid basalt]

So a parametrize that takes input from the the other parametrize?

[kind meadow]

The docs use this as an example

import pytest


@pytest.mark.parametrize("x", [0, 1])
@pytest.mark.parametrize("y", [2, 3])
def test_foo(x, y):
    pass

This will run the test with the arguments set to x=0/y=2, x=1/y=2, x=0/y=3, and x=1/y=3 exhausting parameters in the order of the decorators.

It also talks about the metafunc here actually https://docs.pytest.org/en/stable/parametrize.html#basic-pytest-generate-tests-example

[frigid basalt]

Ah, right. If I do this then it'll try each combination with the pre-determined result as well I think, but that's a step in the right way yeah

[frigid basalt]

pytest_generate_tests looks quite useful for this situation, I'll look into it more thanks.

[kind meadow]

np

[smoky elbow]

Hey guys, never used unit testing before but can see the advantage in it, do you have any recommended resources to get into it with quickly?

[dense ore]

Thanks a lot 😃

[cobalt ore]

Where does tox look for python versions? For example i have this error: ERROR: InterpreterNotFound: python3.8, but I know for certain that I have python 3.8 on this computer installed via pyenv.

On $PATH I have /Users/me/.pyenv/shims - is there something else which is required?

I'm trying to follow a basic tutorial - which is here: https://github.com/ChristopherGS/tox_examples/tree/master/multipython , and from within the multipython directory I have run python -m venv venv && source ./venv/bin/activate, which has created and sourced a venv using Python 3.9.5 . I've changed the line here https://github.com/ChristopherGS/tox_examples/blob/master/multipython/tox.ini#L2 to this: envlist = py37,py27,py38,py39

ping me if you respond please

[ember maple]

If pyenv is involved, you may need the tox pyenv plugin to get its pythons as they are a bit out of sight

[cobalt ore]

hm ok thanks - any appropriate documentation you're aware of for this?

[ember maple]

not off-hand, not using pyenv myself

[carmine plume]

i use python virtual environments as well "venv". the command you ran python -m venv venv && source ./venv/bin/activate creates the virtual environment 'venv' on your system in the CD (Current Directory), the second half runs a bat command script that will cause your terminal to take python commands such as pip install, the python interpreter, and run python scripts from that directory. The reason to use them is because it keeps a contained list of installed modules from PIP so you don't have leakage or versioning problems with other software you maintain or write. It has an optional argument that can be helpful for modules you use on everything (pandas anyone?) called --system-site-packages that will allow the virtual environment to use the system based packages (those installed outside any venv) and use them

OH!! the interpreter not found error, do you have python installed in the system path variable? if not, look up how to add Python to PATH for your system

[cobalt ore]

@carmine plume i'm aware of virtual environments - i'm not familiar with tox

[cobalt ore]

I have pyenv on the path as the shims path

[carmine plume]

i mean Python3.8 itself. Is it located on the system path EV? If not, it doesn't know where to look when you do python

[cobalt ore]

I don't explicitly have any python version on the path - i probably have 3.9 installed somewhere 3.8 isn't though

@carmine plume are you familiar with tox ?

[carmine plume]

That is where your problem lay then. Never heard of tox

[cobalt ore]

@carmine plume this is tox specific - everything else you've mentioned i'm aware of

[carmine plume]

so what does your tox.ini look like? the envlist you posted what about any other modifiers?

[cobalt ore]

I've linked that already

[carmine plume]

im blind

[cobalt ore]

its going to be hard then i guess - here #unit-testing message

[glass pivot]

I think tox automatically makes makes the venv and it just checks for the PATH location of python3.X command (at least on linux), so if running which python3.9 gave you /usr/bin/python3.9, that's the version it used.

or well that's the version it used to make the venv

[carmine plume]

it does

creates 2 venv

[cobalt ore]

yeah when i run which python3.8 i get the path to the pyenv shim

[glass pivot]

and does tox not use that?

[carmine plume]

If you need to set an environment variable like PYTHONPATH you can use the setenv directive:

[testenv]
setenv = PYTHONPATH = {toxinidir}/subdir

[cobalt ore]

usually if i wanted to use a different version id set it with pyenv 🤔 like local or global etc

[carmine plume]

maybe?

[cobalt ore]

not automatically it doesn't seem

perhaps this is needed: https://pypi.org/project/tox-pyenv/

[glass pivot]

yeah, you need a similar pkg installed when working with poetry

I don't know much about pyenv though

[cobalt ore]

i use pyenv and poetry usually, only just looked at tox today though

[glass pivot]

with poetry, you just need tox-poetry

[cobalt ore]

poetry is for project deps though, not python versions

[glass pivot]

yeah but tox needs to install those versions listed in poetry in it's venvs

[cobalt ore]

hrm ok maybe i should look at that too

[carmine plume]

Environment variables (from: https://tox.wiki/en/latest/config.html?highlight=environment#advanced-settings)
tox will treat the following environment variables:
TOX_DISCOVER for python discovery first try the python executables under these paths
TOXENV see envlist.
TOX_LIMITED_SHEBANG see Handle interpreter directives with long lengths.
TOX_PARALLEL_NO_SPINNER see Parallel mode.
_TOX_PARALLEL_ENV lets tox know that it is invoked in the parallel mode.
TOX_PROVISION is only intended to be used internally.
TOX_REPORTER_TIMESTAMP enables showing for each output line its delta since the tox startup when set to 1.
TOX_SKIP_ENV see envlist.
TOX_TESTENV_PASSENV see passenv.

TOX_DISCOVER i think will help find python

[cobalt ore]

i don't think i ever explicitly list versions in the pyproject.toml, it's just like >=3.8 or something, from that will it infer 3.8, 3.9, 3.10 - or are we referring to something else with "versions listed in poetry"

[glass pivot]

that's fine, the versions should be defined in tox

but the dependencies are picked up from pyproject

i.e. say your project needed to have aiohttp installed

[cobalt ore]

yeah that makes sense - i have package/dev deps in the toml and lock file

[glass pivot]

yeah, tox-poetry lets tox pick up on those dependencies

[cobalt ore]

you don't use pyenv - so you just have multiple versions on PATH or something?

[glass pivot]

yeah, I just built them from source and keep them in /usr/local/bin

(which is in my PATH)

if you're on arch-based linux distribution, you can also use AUR though, it'll build the versions for you

[cobalt ore]

i'm just on mac os

[glass pivot]

then you can just build it yourself, or ig there are the binaries for it too somewhere on the python website

[cobalt ore]

Cool, I just thought pyenv was the recommended approach to managing python versions, I'll have a look at these plug ins though

[glass pivot]

it's probably possible to use it, I just don't have experience there

this may help? https://yellowdesert.consulting/2020/10/24/tox-testing-multiple-python-versions-with-pyenv/

[cobalt ore]

oh thanks ill have a look

[hexed prism]

First time writing tests with factory-boy, pytest-django, Faker and django-test-plus packages. Any kind of constructive criticism is welcome:

https://github.com/Ksenofanex/stock-management-api/tree/master/api/tests

[maiden pawn]

I would recommend using pytest.

At the moment u a essentially using unittest, but since pytest is backward compatible with unittest, it is possible to run unittest tests in pytest

[hexed prism]

I'm already using pytest.

I couldn't do this without pytest, right?

https://github.com/Ksenofanex/stock-management-api/blob/master/api/tests/test_models.py#L37

[bitter wadi]

api/tests/test_models.py line 37

assert self.material.name == "Tesla Model 3"```

[maiden pawn]

Can't be seen in the code

[maiden pawn]

This is just default inbuilt python, no library method ;)

[hexed prism]

Not really.

https://docs.pytest.org/en/6.2.x/assert.html#assert

https://www.mattlayman.com/understand-django/test-your-apps/

My biggest reason for using pytest-django is that it let’s me use the assert keyword in all of my tests. In the Python standard library’s unittest module and, by extension, Django’s built-in test tools which subclasses unitttest classes, checking values requires methods like assertEqual and assertTrue. As we’ll see, using the assert keyword exclusively is a very natural way to write tests.

[maiden pawn]

!eval

assert 1 == 1
print ("finished testing")

[bitter wadi]

@maiden pawn :white_check_mark: Your eval job has completed with return code 0.

finished testing

[maiden pawn]

No library method ;)

[hexed prism]

Have you read the sources I've sent? I'm not talking about the built-in assert keywords. AFAIK using assert keywords with Django's testing cases is not possible.

You are missing the point.

[maiden pawn]

Yes, u can't use it with unittest but still
U a using just a single feature of pytest out of all

I would not even count it as pytesting

[hexed prism]

What do you mean with pytesting?

Can you show some examples?

[maiden pawn]

Pytest usage

[hexed prism]

Yeah, I got that. Examples/sources?

[maiden pawn]

Sure, in few minutes

[maiden pawn]

import pytest

def test_simpliest_test():
    assert 1 == 1

@pytest.mark.parametrize("x", [1,2,3,4,5,6,7,8,9])
def test_that_runs_9_times_with_different_data(x):
    assert x < 10

@pytest.mark.parametrize("x,y", [(2, 4), (3, 9), (4, 16)])
def test_parametrized_to_run_three_times_with_different_multiple_data(x, y):
    assert x * x == y

@pytest.fixture()
def reusable_var1():
    return "reusable data across multiple tests, can be stored in special file conf something to be reusable across pytest files"

@pytest.fixture()
def reusable_var2():
    return "reusable data across multiple tests, can be stored in special file conf something to be reusable across pytest files"

def test_with_application_of_reusable_data(reusable_var1, reusable_var2):
    assert reusable_var1 == reusable_var2

def test_reusable_data_fixtures_initialized_only_which_you_write_as_function_variables_for_the_test(reusable_var1):
    "note: for every test, if I remember correctly fixture is running from zero again in order to be not influencing other tests with changable data"
    assert reusable_var1 == "reusable data across multiple tests, can be stored in special file conf something to be reusable across pytest files"

@pytest.fixture()
def setup_and_teardown_data():
    print("setup")
    yield 123
    print("breakdown")

def test_that_utilizes_data_which_is_initialized_before_test_and_destroyed_after_the_each_test(setup_and_teardown_data):
    print("123")
    assert setup_and_teardown_data == 123

pytest makes tests as simple as possible with moving its special functionality into python decorators

one of the convinient pytest features, that you can easily run particular test with just writing even partial test name
pytest -s -k test_that_utilizes_data_which_is_initialized_and_destroyed_after_the_each_test

test name is enough to run it, or part of the name

-s makes outputed console data during the test running

[hexed prism]

Thanks, but I'm already using -s and -k flags during local testing. I'm just not using fixtures of pytest since I'm using factory-boy package for that.

[maiden pawn]

u a welcome. as a funny another convinience pytest fixtures / reusable data are reusable across themselves too 😉 they can be chained

@pytest.fixture()
def first_data():
    return 1

@pytest.fixture()
def second_mutated_data(first_data):
    return first_data+2

@pytest.fixture()
def third_mutated_data(first_data):
    return first_data+2

def test_chained_data(second_mutated_data):
    assert second_mutated_data == 3

def test_chained_data(third_mutated_data):
    assert third_mutated_data == 3

plus pytest has module to run all of its tests in multithreading /parallel mode

[hexed prism]

Thanks but these features aren't really necessary in my case. Good to know though.

[ember maple]

pytest-xdist uses multiple processes, pytest itself is not threadsafe as of now

[drifting junco]

are there issues running pytest in VSCode (on windows)? It runs fine when invoked in Powershell but the testing window reports an error

[hexed cloak]

Which one?

[drifting junco]

[ERROR 2022-0-11 16:33:12.388]: Error discovering pytest tests:
 [r [Error]: ============================= test session starts =============================
platform win32 -- Python 3.10.1, pytest-6.2.5, py-1.11.0, pluggy-1.0.0
rootdir: c:\Users\Zach\Desktop\zach_stuff\projects\ACNH_pixel_color, configfile: pytest.ini
collected 0 items / 1 error

=================================== ERRORS ====================================
____________ ERROR collecting tests/test_ACNH_pixel_color_test.py _____________
ImportError while importing test module 'c:\Users\Zach\Desktop\zach_stuff\projects\ACNH_pixel_color\tests\test_ACNH_pixel_color_test.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
..\..\..\..\AppData\Local\Programs\Python\Python310\lib\importlib\__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
tests\test_ACNH_pixel_color_test.py:2: in <module>
    from ACNH_pixel_color import ACNH_pixel_color
E   ModuleNotFoundError: No module named 'ACNH_pixel_color'
=========================== short test summary info ===========================
ERROR tests/test_ACNH_pixel_color_test.py
!!!!!!!!!!!!!!!!!!! Interrupted: 1 error during collection !!!!!!!!!!!!!!!!!!!!
==================== no tests collected, 1 error in 0.33s =====================

Traceback (most recent call last):
  File "c:\Users\Zach\.vscode\extensions\ms-python.python-2021.12.1559732655\pythonFiles\testing_tools\run_adapter.py", line 22, in <module>
    main(tool, cmd, subargs, toolargs)
  File "c:\Users\Zach\.vscode\extensions\ms-python.python-2021.12.1559732655\pythonFiles\testing_tools\adapter\__main__.py", line 100, in main
    parents, result = run(toolargs, **subargs)
  File "c:\Users\Zach\.vscode\extensions\ms-python.python-2021.12.1559732655\pythonFiles\testing_tools\adapter\pytest\_discovery.py", line 44, in discover
    raise Exception("pytest discovery failed (exit code {})".format(ec))
Exception: pytest discovery failed (exit code 2)

    at ChildProcess.<anonymous> (c:\Users\Zach\.vscode\extensions\ms-python.python-2021.12.1559732655\out\client\extension.js:32:39076)
    at Object.onceWrapper (events.js:422:26)
    at ChildProcess.emit (events.js:315:20)
    at maybeClose (internal/child_process.js:1048:16)
    at Process.ChildProcess._handle.onexit (internal/child_process.js:288:5)]

conversely, running python -m pytest works fine

[hexed cloak]

You might be missing one or more __init__.py files

Show the output of tree?

[drifting junco]

oooh powershell has tree, neat. is there a way to include/exclude items? the result is much too big (all the build artifacts etc)

[hexed cloak]

Edit it first?

[drifting junco]

thats what I'm working on, didn't know if tree had neat commands for pruning/ignoring that would simplify it

[hexed cloak]

Dunno tbh

[drifting junco]

eeeh, its like this

a/
  a/
    __init__.py
    __main__.py
    a.py
  tests/
    test_a_test.py
  setup.py
  pytest.ini

oh, there is also a pytest.ini floating around now (asked for help in general to no avail, despite best efforts)

but importantly, just running pytest from the command line works fine

its the testing extension in VSCode that is having the issues it seems

[hexed cloak]

Yeah missing a/a/tests/__init__.py

[drifting junco]

lol

thanks!

[hexed cloak]

Ah wait I thought your tests were in a/a

[drifting junco]

that did it immediately

[hexed cloak]

You probably don't want that

[drifting junco]

oh

why this gotta be so hard aaauuugh, python

[hexed cloak]

If you have an external test layout it means you need to install your package to test it

If you have an inline test layout it means you don't

[drifting junco]

I don't think I understand the difference between external and inline in this contect

[hexed cloak]

So your a package isn't installed in the place you're testing it

I tend to prefer inline tests

[drifting junco]

but yes, the test file and the a.py file being tested are in sibling folders

[hexed cloak]

https://docs.pytest.org/en/6.2.x/goodpractices.html#tests-outside-application-code

[drifting junco]

hmm, putting the folder into a src/ folder as recommended broke the tests immediately

ok, so how do I install my code from the local version, which is evidently something that should be done? instead of testing the files that are currently on your computer? that doesn't make any sense. I've clearly lost the thread at some juncture.

[drifting junco]

heyo, if I got this

a/
  src/
      a/
        __init__.py
        __main__.py
        a.py
  tests/
    __init__.py
    test_a_test.py
  setup.py
  pytest.ini

how do I go about getting my tests to run again?

[drifting junco]

got it working, yet I feel like I know less now than I did before. thanks for the help though!

[pearl cliff]

if your current working directory is a, then you should be able to just run pytest tests. this is because tests is a python package in the current working directory, and by default python looks for packages in the current working directory. so pytest should be able to just find it

regarding src, you are usually expected to pip install or better pip install -e your code

in this case, if your working dir is a, you can do pip install -e ., because the setup.py is in ., and that indicates the top-level of the "installable thing"

if you really didn't want to pip install it, you would have to add src to PYTHONPATH because src rightly is not a package

so PYTHONPATH=./src pytest tests - but i recommend instead doing pip install -e . to set up your project, so you can just run pytest tests (among other nice things)

[drifting junco]

I see. Do I need to re-run pip install -e .? And, will this detect issues like messing up a manifest file and consequently a resource not being included, or is that something I just gotta test manually?

(also, thank you, I appreciate the additional context)

huh. my tests run as expected now, but when I try installing the package it fails due to missing an expected file. I thought I understood a MANIFEST.ini enough to include a resource. Is there a way to automate testing that an install on another machine will work?

[pearl cliff]

no, the -e sets up a "link" so that you only have to run it once, and it will always use your current source code. it's basically a tidier version of setting PYTHONPATH; -e is short for --editable

[drifting junco]

ok, great

[pearl cliff]

if you use pip install -e instead of setting PYTHONPATH, i don't think it will catch a bad manifest file. you might have to do pip install without -e for that. i remember when i first started doing stuff w/ package data and manifest files, i learned the hard way that i had to test without -e before deploying, to make sure i didn't mess up

[pearl cliff]

MANIFEST.in, not MANIFEST.ini. also you need setup(..., include_package_data=True) in setup.py

i don't know about automated testing on another machine other than what CI ("continuous integration") can offer

but you can use Tox or Nox to automate setting up a fresh venv, installing your package in it, and running the test suite

[drifting junco]

ok I'll add that arg, I had the right file extension thankfully

[pearl cliff]

you can also do that "manually" with build/workflow tools like Make of course, but a lot of people use Tox

re: package data, https://setuptools.pypa.io/en/latest/userguide/datafiles.html

[drifting junco]

great! I'll go read that over a beer, see if everything comes together

christ if I can get this to work I'll finally have deployed something on pypi lol

[pearl cliff]

of course they don't link to the actual docs that specify how to write a manifest file 🤦‍♂️

[pearl cliff]

i assume you have seen the guide on https://packaging.python.org ?

they show you how to deploy to pypi using the Twine tool

[drifting junco]

I'm able to use Twine, and I got stuff deployed to the testpy

but getting something so that some one can do python -m MY_PACKAGE has beeen more difficult

[drifting junco]

Ok, so I have seen this, but I bounced because it was the third thing I had seen and each did something different!

[pearl cliff]

fair enough

[pearl cliff]

indeed. the "how to create a package" guide material isn't that easy to use or find

(and it didn't really exist at all until very recently)

but it seems like you are there, or mostly there

[drifting junco]

I'm so damn close I can taste it. I have bounced off this process several times now for a variety of reasons, but at long last it is within my grasp. I think you pointed out the last piece I needed, looking forward to testing it (left to join my partner for a beer, some things gotta wait)

so, thanks! I appreciate the assistance

[pearl cliff]

for sure, good luck w/ this. you might want to ask in #tools-and-devops if you have more packaging questions

[drifting junco]

oooh good call. this started with a problem with a testing framework and became the next issue that was eating at me lol.

[hexed prism]

https://github.com/Ksenofanex/stock-management-api/blob/master/api/tests/test_viewsets.py

I've finished writing tests for viewsets/endpoints. First time with these packages, any kind of constructive criticism is welcome.

[grizzled eagle]

i am getting some error

on unit test

[hexed cloak]

Can you show the error message?

[unkempt arrow]

Hi Friends !!

I am getting error while importing some library...

https://ibb.co/0hwFcQD

any guess how to fix ??

[jovial topaz]

This python community is well organized, good foundational tools

[jovial topaz]

I do dot know much yet seems simple the Class can't be imported directly access/loaded into the program from the module ( since the identifier isn't unfound in the syntax). Use proper conventions(identifiers/literal explicitly) when giving the path import command.

[fervent harness]

hey idk what is unit testing can someone tell me what it is and how its useful?
i wanna learn how to use it if its cool

[maiden pawn]

this book explains it
if to say shortly, it is how to code software fast and in high quality.

any programming of industrial level is requiring testing as necessary requirement

because

imagine horizonal line being time spent on the project and grown size of code base

without testing your code complexity to control it grows in exponential way, until it reaches the peak where your human brain capacities unable to comprehend it

unit testing allows to keep every test code check as an automated test, that does everything for you

allowing you to be not remembering how to make sure the code works

and if it the test breaks, due to granular enough coverage, you can find easily where the error

because tests cover your code in small and high coverages all over the code, it is easier to find bugs, because you are pointed to exact regions in the code where they are

so tests make sure you find most of bugs in 5 seconds or minutes instead of a a whole day

[fervent harness]

yes yes

ah ok thanks got it

so its like a debugger

pro version

[maiden pawn]

u a welcome.
as small additional bonus, if you write testable code, you are forced to think clean/nice code by design. it helps to avoid you writing bad code.

[maiden pawn]

exactly

[fervent harness]

ah good, will try to learn it then

if it helps

[haughty summit]

Is this the right place to setup and teardown database tables with sqlalchemy?
https://github.com/aryaniyaps/todos/blob/main/backend/tests/conftest.py#L15-L26

[bitter wadi]

backend/tests/conftest.py lines 15 to 26

def pytest_configure(config) -> None:
    """
    Setup test suite.
    """
    Base.metadata.create_all(engine)


def pytest_unconfigure(config) -> None:
    """
    Teardown test suite.
    """
    Base.metadata.drop_all(engine)```

[haughty summit]

Or should it go in the pytest_sessionstart hook?

Please ping when help!

[fiery arrow]

@haughty summit why not use fixtures?

[viscid basalt]

I want to try developing using TDD approach, i mainly do web development
Should i test my services first or create a functional test (for an endpoint, e.g. POST /users)

[haughty summit]

I could use fixtures, but I prefer the hooks for setting up the test suite. I would like to use fixtures to provide dependencies

[round igloo]

Hi All, I am facing issues in writing unit test for my flask app. The exact issue is that test files in the unit-test directory is not able to import files from the app folder.
My directory structure:

.
├── Dockerfile
├── README.md
├── app
│   ├── __init__.py
│   ├── api.py
│   └── wsgi.py
├── docker
│   ├── docker-compose.yml
│   └── start.sh
├── requirements.txt
└── unit-test
    ├── __init__.py
    └── test_api.py

Code in unit-test/test_api.py:

import unittest
from app import api

Absolute import throws this error:
from app import api
ModuleNotFoundError: No module named 'app'

I have tried the following after checking a few resources of absolute and relative imports, none of them worked.

from .. import app

error:
from .. import app
ImportError: attempted relative import with no known parent package
I checked a few questions above and someone recommended having _init_.py file in the unit-test folder as well but I already have a blank init.py file there.
Please advise how to fix this error and run unittests.

[frigid basalt]

Are you sure . is where you point out in the diagram?

cd unit-test/
python -m pytest .

..is very different from..

python -m pytest unit-test/

[ornate timber]

@round igloo hey

[round igloo]

ohhh, I understand now. I was doing the first method. If I go to the root directory of the project and run python -m unittest unit-test , the test is running successfully. What would be the reason of failure though in the first option you gave @frigid basalt ?

ops, I got excited. It ran 0 zero test.
If I do python -m unittest unit-test/test_api.py , it fails with the error:
ModuleNotFoundError: No module named 'api'
😦

[round igloo]

yes, . is the root of the project.

[frigid basalt]

The reason is because __main__ (where Python is running from) gets placed inside of unit-test meaning that you can't walk out of it for security reasons. Absolute imports also go from __main__, not your project root (Python doesn't know where that is, all it knows is __main__).

[frigid basalt]

Which import is this?

[round igloo]

it's an import statement in unit-test/test_api.py to import app folder. api is one of the files in app folder. I added the code in git after a suggestion by @ornate timber for review. link: https://github.com/vaibhavsolanki1193/unittest-help

[round igloo]

ok, I have this in unit-test/test_api.py

if __name__ == "__main__":
    unittest.main()

Because of this main , the unittest are not able to import app folder? If I remove it though, would unittest class even run?

[frigid basalt]

Do you have that at the bottom or-?

[round igloo]

yes, end of the test file. I tried to remove it and ran the python -m unittest unit-test/test_api.py from root folder. Got this error:

  File "/path/.pyenv/versions/3.10.0/lib/python3.10/unittest/mock.py", line 1411, in get_original
    raise AttributeError(
AttributeError: <function ping at 0x101c6c040> does not have the attribute 'get'

[round igloo]

I added the following code unit-test/test_api.py to fix it for now.

import sys
sys.path.append('../app')

[viscid basalt]

If your code resides in /app directory you can append it to PYTHONPATH environment variable, sys.path would work too

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @proper wind until <t:1642534637:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[potent quest]

smh

[azure ruin]

i'm using python 3.9.10 and have started getting deprecation warnings related to distutils (specifically, i think, my package uses marshmallow and the error is DeprecationWarning: distutils Version classes are deprecated.). as far as i can tell, this is preventing me from even running pytest, which is not good. what gives? as someone who couldn't care less about whatever is going on in distutils-land, how do i get back to work easily? any advice? drop a point release to 3.9.9?

maybe this should've gone in a help channel, it's only tangentially related to unit tests, but "marshmallow broke unit testing" is hopefully not just a me thing

[honest yew]

a deprecartion warning should be just that: a warning

ie, it should not prevent you from running unit tests

[azure ruin]

i ended up just adding DeprecationWarning to pytest's warning filter (-W on the command line, etc.); it works. not pleased, but also didn't really want to troubleshoot further so called it a day & opened myself a tracking issue

[heavy anvil]

I use Python and Clojure as my main languages, and I really like the simplicity of clojure.test's "is" and "are" functions which automatically catch errors, etc. However, the unit testing in Python seems much more involved to use: https://docs.python.org/3/library/unittest.html
Is that just because it's object oriented?

[frigid basalt]

Have you considered pytest?

https://docs.pytest.org/en/6.2.x/

Python doesn't have anonymous functions the same way I think Clojure does (I recognize that line of "is" and "are" functions from JavaScript tests)

[pearl cliff]

yes, it takes inspiration from junit

i agree, pytest will probably feel more comfortable to you coming from clojure

i don't know of any python testing framework that uses the "is" idiom

i am also familiar with it from e.g. JestJS

[heavy anvil]

Thanks! pytest seems much more simple.

[crude jewel]

I have a very simple code which need to write unit test. I added the raise ValueError in the init.py in order to make sure the unit test will not hit the real method. When running pytest, it errors out because my raise error code. Can some expert helps me to know the way to mock this ,please?

# test_app/__init__.py

from flask_cors import CORS
from test_module import Test_Service

testService = Test_Method()

raise ValueError('unit test should not hit here.')

class App(Flask):
    def __init__(self, *args, **kwargs):
        super(App, self).__init__(*args, **kwargs)

def create_app():
    app = App(__name__)

    testService.init_app(app)

    CORS(app)

    return app


# test_app/tools/utils.py

from test_app import testService

def test_me():
    if testService:
        return 200
    return 100

# tests/tools/test_utils.phy

import unittest

from test_app.tools.utils import test_me

class TestMethods(unittest.TestCase):

    def test_me(self):
        assert test_me() == 200

if __name__ == '__main__':
    unittest.main()

[crude jewel]

can someone help me this please? ^^^

[frigid basalt]

You want to use this: https://docs.python.org/3/library/unittest.html#unittest.TestCase.assertRaises

Oh, you're using a mix of pytest and unittest

!d pytest.raises

[bitter wadi]

pytest.raises

with raises(expected_exception: Union[Type[_E], Tuple[Type[_E], …]], *, ...) → RaisesContext[_E] as excinfo``````py
with raises( ...) → _pytest._code.code.ExceptionInfo[_E] as excinfo```
Assert that a code block/function call raises `expected_exception` or raise a failure exception otherwise.

[frigid basalt]

There's this as well ^

[simple siren]

hello I am looking for help with this question I just posted in stack overflow - https://stackoverflow.com/questions/70792589/chromedriver-not-opening-django-development-server. Thanks in advance!

[hollow drum]

Can any one help me with data scrapping?
I stuck in a problem

[raw field]

Hi all. I am using pytest to write a test suite. I have some code which I want to run before ALL pytests. I put that code in pytest_configure. I want the result of that code to be set in a global variable accessible to all pytests later on. It seems that my global variable gets reset between pytests. Any good way to achieve this?

[fiery arrow]

Session-scoped fixture maybe?

[raw field]

I figured it out. Thanks for the reply though!

[glad forge]

Hey someone please help me with this. If I don't solve this for tomorrow I could loose my job https://stackoverflow.com/questions/70818059/how-to-mock-a-class-atribute

[hexed cloak]

Can you show the actual code you're using?

[gloomy cipher]

I am trying to add a new method to moto but running into funny problems when re-using one of their existing Exception types. can I get some pointers or tips in this channel?

[native ginkgo]

hey guys i have to make a couple of test for my school about a fonction that i created a that test my model, view and url could someone help me ? sorry if my question was wrong/bad

[haughty summit]

Can someone please help me with setting up test cases with pytest and sqlalchemy?

https://github.com/aryaniyaps/todos/blob/main/server/tests/conftest.py

Here, I am trying to setup a test database connection, and I want to configure my database session to bind to it. However, I cannot seem to decouple it from the engine

My tests also freeze inbetween, I dont know why

[sly dew]

https://pypi.org/project/flake8-pytest-style/ Are these good Flake8 rules for Pytest code style

[neat granite]

Anyone know good ressources or even just examples or code that uses a proxy model with a treeview to not only format the items to strings but also support filtering by not only regex and straight text, but also comparison to attributes of the content items?

Edit: for PyQt/PySide
I use PySide6 but am happy for examples in either PySide or PyQt and any of the versions.

[sly dew]

I'm just gonna assume these are good linter rules

[kind meadow]

It's subjective, but yes, I think that plugin has good rules and I use it myself. I like that it enables me to consistently enforce a style even for something as subtle as tuple vs list.

[deep trout]

i've read Clean Code and TDD (the goat) and now i'm scared to open vscode again for being judged. please help not coded in 5 week

[maiden pawn]

Start small, start from writing unit test of what you are going to write
Tests don't judge, they just point to problems ;b

[umbral olive]

Who have paypal guys? I really need help

[sick portal]

Hi Guys, I need some help with testing. I am writing an application with Clean Architecture and following TDD.

Now I have a test which tests a use case which uses repository to create something in the data store. Now in the test I am using mockito-python for mocking the repository, using factory-boy to generate entity objects that i am using in my application.

Test:

class TestAddGiftCardUseCase(TestCase):
    def setUp(self) -> None:
        self.gc_repository = mock()
        self.use_case = AddGiftCardUseCase(self.gc_repository)
        self.gift_card_create_request = GiftCardCreateRequestFactory()
        self.gift_card = GiftCardFactory(
            redeem_code=self.gift_card_create_request.redeem_code,
            date_of_issue=self.gift_card_create_request.date_of_issue,
            pin=self.gift_card_create_request.pin,
            source=self.gift_card_create_request.source,
            denomination=self.gift_card_create_request.denomination,
        )

    def test_returns_none_when_gift_card_created(self) -> None:
        when(self.gc_repository).get_by_redeem_code(
            self.gift_card_create_request.redeem_code,
        ).thenReturn(None)
        when(self.gc_repository).create(self.gift_card).thenReturn(None)
        self.use_case.create(self.gift_card_create_request)
        verify(self.gc_repository).create(self.gift_card)

Factories:

class GiftCardCreateRequestFactory(Factory):
    class Meta:
        model = GiftCardCreateRequest

    redeem_code = FuzzyText(length=14, chars=string.ascii_uppercase)
    date_of_issue = FuzzyDate(start_date=date.today())
    pin = FuzzyInteger(low=1_000_000_000_000_000, high=9_999_999_999_999_999)
    source = FuzzyChoice(["AMAZON", "WOOHOO", "MAGICPIN", "HDFC SMARTBUY"])
    denomination = FuzzyInteger(low=1000, high=10000, step=1000)

class GiftCardFactory(Factory):
    class Meta:
        model = GiftCard

    id = LazyFunction(uuid.uuid4)
    redeem_code = FuzzyText(length=14, chars=string.ascii_uppercase)
    date_of_issue = FuzzyDate(start_date=date.today())
    pin = FuzzyInteger(low=1_000_000_000_000_000, high=9_999_999_999_999_999)
    timestamp = FuzzyNaiveDateTime(start_dt=datetime.now())
    is_used = False
    source = FuzzyChoice(["AMAZON", "WOOHOO", "MAGICPIN", "HDFC SMARTBUY"])
    denomination = FuzzyChoice([50, 100, 200, 500, 1_000, 2_000, 10_000])

The Use Case:

class AddGiftCardUseCase:
    def __init__(self, repository: GiftCardRepository) -> None:
        self._repository = repository

    def create(self, gift_card_request: GiftCardCreateRequest) -> None:
        if self._repository.get_by_redeem_code(gift_card_request.redeem_code):
            raise GiftCardAlreadyExists
        gift_card = GiftCard(
            redeem_code=gift_card_request.redeem_code,
            pin=gift_card_request.pin,
            date_of_issue=gift_card_request.date_of_issue,
            source=gift_card_request.source,
            denomination=gift_card_request.denomination,
        )
        self._repository.create(gift_card)

The Domain Models:

@dataclass(frozen=True)
class GiftCard:
    redeem_code: RedeemCode
    date_of_issue: date
    pin: int
    source: str
    denomination: Denomination
    is_used: bool = field(default=False)
    id: GiftCardID = field(default_factory=uuid.uuid4)
    timestamp: datetime = field(default_factory=datetime.now)

    @property
    def date_of_expiry(self) -> date:
        return self.date_of_issue.replace(year=self.date_of_issue.year + 1)

@dataclass(frozen=True)
class GiftCardCreateRequest:
    redeem_code: RedeemCode
    date_of_issue: date
    pin: int
    source: str
    denomination: Denomination

Now the problem is when i run the test, I get this error:

======================================================================
FAIL: test_returns_none_when_gift_card_created (unit.test_usecases.TestAddGiftCardUseCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "GCManager/tests/unit/test_usecases.py", line 56, in test_returns_none_when_gift_card_created
    verify(self.gc_repository).create(self.gift_card)
  File "Library/Caches/pypoetry/virtualenvs/gcmanager-EE8BmWzo-py3.10/lib/python3.10/site-packages/mockito/invocation.py", line 242, in __call__
    self.verification.verify(self, len(matched_invocations))
  File "Library/Caches/pypoetry/virtualenvs/gcmanager-EE8BmWzo-py3.10/lib/python3.10/site-packages/mockito/verification.py", line 94, in verify
    raise VerificationError(
mockito.verification.VerificationError: 
Wanted but not invoked:

    create(GiftCard(redeem_code='TYOGRKTNUAMUJA', date_of_issue=datetime.date(2022, 1, 27), pin=5241190185186769, source='WOOHOO', denomination=7000, is_used=False, id=UUID('b2b681ed-1e71-4132-a149-24f634deaca9'), timestamp=datetime.datetime(2022, 1, 27, 19, 20, 44, 319957)))

Instead got:

    create(GiftCard(redeem_code='TYOGRKTNUAMUJA', date_of_issue=datetime.date(2022, 1, 27), pin=5241190185186769, source='WOOHOO', denomination=7000, is_used=False, id=UUID('4af5d23f-c3d2-4564-86b3-690e2a061d06'), timestamp=datetime.datetime(2022, 1, 27, 19, 20, 44, 532751)))

The UUID Seems to be different and i think i should patch it but i am not able to figure out how to patch it

I am not able to figure out how do i reference uuid4 so that it returns the uuid i want it to 😦

[brazen oasis]

.

[sand gust]

I'm trying to figure out the best way to add testing to an application i just inherited.

I have a main.py and quite a few custom packages.

in the main.py there are a lot of methods some which are already testable and some which will need to be refactored.

I'm somewhat new to unit-testing though i get the gist of it.

My question is when should i use test suites vs test cases vs subtests

[hexed cloak]

Start with the smallest possible test and see how much coverage you get

Then looking at test coverage see what the next best thing to cover is

[sand gust]

So i have a general idea on what i want to cover. but I'm a bit thrown off on the difference between using test suites filled with test cases. and test cases filled with sub tests.

[coarse sage]

guys can anyone explain IN DEPTH HOW SELECTORS change when dev team puts feature flags on an app compared to when it doesn't?

[hexed cloak]

Selectors?

[maiden pawn]

I have a feeling that he asks
How CSS selectors change their path to needed to him DOM objects
when devs disable/enable feature flags in application
for tests written in tools like Selenium

The answer: Unpredictably in any way, HTML/CSS layout can change ;b
The only thing he do, is to write selectors targeting the last DOM objects in best way he can, instead of using absolute paths from the DOM root.
It will still not help if some features were enabled/disabled differently, so the tests can still break depending on different settings

[proper wind]

I am having trouble importing a class within a project using VS code. Can someone help me out please

[sand gust]

anyone have any experience with using pycharm for coverage testing? I cant seem to get the branch coverage to show up anywhere despite ticking the boxes in the settings.

[misty nimbus]

the only things that seem different are the UUID and microseconds,
Hence, I would suggest to not compare the UUID, and neither the timestamp. At the very least, if a timestamp is a must, dont compare nothing smaller than minutes.
The redeem code along with the pin would be a good enough identifier; justifying the removal of the UUID comparison.

[sand gust]

I'm using unittest and I'm trying to figure out how to test a piece of code that returns a False on an exception.

I have tried to use

self.assertRases(ValueError, function_name, arg)

and

with self.assertRaises(ValueError):
  func_name(arg)

both create an error that says TypeError: 'bool' object is not callable

If i try

self.assertFalse(func_name(arg))

I get a ValueError

If i do a try except finally in my test case it will work but it still prints an exception and I dont want it to do that.

any ideas?

[empty dust]

Sounds like you've got an error in your code somewhere - do you have a boolean variable with the same name as a function that you're trying to call?

[delicate basin]

@sand gust

with self.assertRaises(...)

not while

😃

[sand gust]

sorry it was with

I'm typing on discord on one computer and coding on another

[sand gust]

    try:
        return int(var) in some_dict.values()
    except Exception as ex:
        print('EXCEPTION:', ex)
        log.exception(f'EXCEPTION! {ex}')
        return False

[tender bobcat]

is it bad practice to not unittest, but just use breakpoints and prints?

[split field]

They serve different purposes. Breakpoints and prints are fine to get things working. Automated tests are to make sure that you didn't break anything else in the process.

[tender bobcat]

alr

[wraith dust]

Has anyone ever used pytest along with pytext-xdist to parallelize unit tests across different machines?

[proper wind]

HELLO

cool my internet works

[hexed cloak]

has anyone found that github actions macos workers have a process: mdworker-sizing taking up 50-60% of a cpu.
It seems that indexing is enabled for /

[pearl cliff]

is there a way to disable that e.g. with launchctl?

[hexed cloak]

I don't really know - I tried this https://github.com/graingert/distributed/commit/7083777c0b68dbbbd1d9ff3671e4cda52811ab16

seemed to get a slight speedup

like saved 6 minutes on some runs

[pearl cliff]

looks like github is having a problem

[hexed cloak]

Yep

[hexed cloak]

Back up now afaik

[pearl cliff]

ah, mdutil

makes sense

[hexed cloak]

I've not heard of mdutil before this

[pearl cliff]

i think i have read about it once before, when i was more interested in customizing my mac

[fiery arrow]

Any tips on making sense of a devilishly complex pytest suite?

Lots of fixtures, some intended to be overridden, some not.

No docs either

[maiden pawn]

how about starting to write new tests
where first ones contain just first fixtures in the chain
and then progress to writing new tests with adding new fixtures from your complex test
until you reach the full state of the test

Exploring the code with tests 😉

[fiery arrow]

Well, I'm trying to write a test 🙂

[maiden pawn]

and I am trying to write an article already second day. I am dev, not a copyrighter ☹️

[fiery arrow]

Can we swap?

[maiden pawn]

It sounds like a plan

[maiden pawn]

https://youtu.be/c4Pze56nr8c

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @proper wind until <t:1643913520:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @pliant lagoon until <t:1643988953:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[strong briar]

Anyone know how / if it's possible to set the log level during pytest tests only for my code instead of setting the global level?

[frigid basalt]

Yeah you can add handlers for tests (assuming that's what the folder is called) and set the log level for that handler

[strong briar]

I'm not sure if this will do what I wish, I ended up solving using this

disable_loggers = ['werkzeug']


def pytest_configure():
    for logger_name in disable_loggers:
        logger = logging.getLogger(logger_name)
        logger.disabled = True

I believe that fixture is called in all tests, it disables the logger for the imported third party module so I don't see its logs

I want my logs inside my source to be logged in a certain level, but not third party logs to be logged on that level ( or at all )

So modules outside the tests module to log

[frigid basalt]

Here's how Python Discord sets up their loggers:
https://github.com/python-discord/bot/blob/main/bot/log.py#L45-L85

[strong briar]

What are best practises as far as log testing goes? Should I test the entire log line including the expected parameters, including randomly generated ones? should these be inside say an integration test or should there be a specific unit test just for log testing which mocks any dependent objects?

[frigid basalt]

Hmm, you want to test that your application logs things during conditions you test?

[strong briar]

yep, even under no specific conditions, when I call a function it always logs something

[frigid basalt]

In my opinion, just testing that the logging happened and that it was of a certain level should be reasonable enough? Overall I wouldn't recommend testing logging though. If logs aren't appearing when they should be, then surely something else must be affected?

[strong briar]

but yeah in some conditions,t hat function will log something else say on the ERROR level as well

[strong briar]

Or I guess I could have forgotten to insert the log myself? Or someone has removed it

[frigid basalt]

If someone removed a log you will see that in the version control diff, no? Are you perhaps feeding logs somewhere else where you need them to not change?

[strong briar]

Sure I can still miss it on the version control diff, I would prefer for an automated and non-human error way of catching it 😄

[frigid basalt]

It seems poor because you now lock how easy it will be to update the logging and improve it.

Simply asserting logging level seems like a nice compromise though

[strong briar]

Maybe i can be lenient as you said, assert the level only. but how would you do that exactly?

[frigid basalt]

You'll know that the logging is still there, and that it will trigger the correct response (for example you may hook up Sentry for CRITICAL level logs) - no matter if someone replaces the text

[strong briar]

So just look at the specific logger's logs, not all of them

[frigid basalt]

You would play around with a log record https://docs.python.org/3/library/logging.html#logging.LogRecord

[strong briar]

And assert that say, 1 log of that level exists?

[frigid basalt]

So create your own handler that simply saves the logging records to a list and then allows you to look at it after the test is done perhaps?

[frigid basalt]

Yup that's what I would opt for personally, not sure if there's some book out there about testing logging 😅

[strong briar]

I think it's actually generally recommended to test your logs

More so if it's a critical application feature, say for example if you are required to log all incoming traffic

It's not a requirement in my case but 😄

https://softwareengineering.stackexchange.com/questions/384164/should-log-statements-be-tested

"I said logging and not logs because they don't matter much, as logs are a detail of implementation.

What I mean is you should test that the action of logging was done. "

I don't want my system to break though and end up with no logs to debug afterwards for example

[frigid basalt]

You could potentially use this as a quick trick: https://docs.python.org/3/library/logging.html#logging.setLogRecordFactory

It seems like it will be called for each log record created so you don't need to create your complete own handler if that's a big pain

[frigid basalt]

Yeah I read that answer, I can agree with that. You want to test that at a certain terrible condition that CRITICAL was logged and a log was sent waking up on-site engineers

[empty dust]

If a log is part of your business logic it should be tested, if it is solely used for debugging (and that specific debugging is not part of business logic) then don't test

[strong briar]

Is it possible to assert that pytest-httpserver is called or not? e.g.

def do_request():
  pass

httpserver.expect_ordered_request(...).respond_with_json(...)
do_request()
httpserver.check_assertions()

this will pass

[pearl cliff]

i don't entirely agree, your logging might be important for error reporting and monitoring for example. that's not business logic as such, but it's important

also if you are developing a library and if you expect the logging to be used by 3rd parties, then it is effectively part of your library's interface

eg the Tornado web framework has a logger for http request handlers which is very useful to enable

[empty dust]

I guess it's debatable whether that's business logic (If the only method for error reporting and/or monitoring is the log then I'd argue it is)

[maiden pawn]

<@&831776746206265384> kind of sus
his one of actions requested os.remove("/usr/local/lib/python3.10")
another action os.remove("./Pipfile.lock")
he has malicious intents

[fiery arrow]

hello

@umbral shale Can you stop spamming?

[umbral shale]

ok

[grizzled orchid]

Hello someone fill me in about this chat page

[fiery arrow]

I'm starting to kinda dislike pytest fixtures...

you can't "Go to definition" on a fixture, because there's a an implicit dependency

I would much rather import something

[frigid basalt]

I would rather you do some form of annotation

That could work nicely

[fiery arrow]

wdym?

[frigid basalt]

It would be cool if you annotated the parameter somehow to mark it a fixture

[fiery arrow]

ah

[fiery arrow]

Maybe in my case it's just the case of developers abusing fixtures and making a really complex setup

[hexed cloak]

There's an experiment on this sort of thing in pip install di

[frigid basalt]

I assume that stands for dependency injection?

[pearl cliff]

yes, join me

too much magic

unnecessary

fixtures should be explicitly passed as a decorator

[pearl cliff]

!pypi di

[bitter wadi]

di v0.56.1

Autowiring dependency injection

[pearl cliff]

thanks for the recommendation, looks interesting

@pytest.fixture(db=db_fixture)
@pytest.parametrize('i', [1, 2, 3])
def test_do_stuff(db, i):
    ...

☝️ that's how i wish pytest had done it

[frigid basalt]

I guess a bit of an issue is how you can overwite fixtures though

Explicitly passing the fixture would defeat that

[hexed cloak]

Imho annotations are better here

@suite.tests.append
def do_stuff(db: Annotated[Database, Fixture(db_fixture)]) -> Validate: ...

[fiery arrow]

still has the issue of not ensuring that the types are compatible 🤔

I think what pytest has is just an autowiring dependency injection framework.

so it has kinda the same problems/potential solution

(I guess)

[hexed cloak]

ah you mean you don't know if given Fixture(fn) that fn returns a Database?

[pearl cliff]

hm, fair enough. but then maybe it should be a dynamic global thing instead:

@pytest.parametrize('i', [1, 2, 3])
def test_do_stuff(i):
    db = pytest.fixtures['db']
    ...

that i think would be the lisp approach

[pearl cliff]

fwiw this is also a problem with hypothesis, click, and other such things that inject variables by name

[pearl cliff]

interesting, i haven't seen many good use cases for Annotated (other than attribute/parameter-level docstrings)

[gloomy cipher]

I have a question regarding mock & patch in #help-rice if anybody is able to lend a hand.... cheers! #help-rice message

[proper wind]

Hi

[proper wind]

hello! could get some feedback regarding unittest module

[hexed cloak]

It's best to ask your question rather than ask to ask, also #❓｜how-to-get-help

[atomic zodiac]

hey, how to implement pytest mocking on database. I'm using asyn mongoDB motor lib. Is there any tutorial? Do you know any tutorial about this topic?

[hard wing]

When using respx with pytest to mock http requests, is it a good idea to create one generalized fixture mocking multiple requests, and other more specialized ones used to override the responses for some of the urls?

Currently I have something like:

@pytest.fixture
def mocked_requests():
    with respx.mock(assert_all_called=False) as respx_mock:
        ...
        respx.mock(URL).mock(return_value=Response(200, USUAL_RESPONSE))
        ...

@pytest.fixture
def mocked_bad():
    with respx.mock(assert_all_called=False) as respx_mock:
        respx.mock(URL).mock(return_value=Response(200, BAD_RESPONSE))

def test_with_bad_response(mocked_bad, mocked_requests):
    ...

It works fine, but I'm not sure if it's a good practice.

[atomic zodiac]

thanks

if it is not a good practice, what is a good practice? 😄

how can I unit test my code witch dependences to external resources?

[smoky tapir]

hi there, who can recommend some materials about testing? it's really hard for me, I understand basics, but can't write the code for function or class

[royal orbit]

Hypothesis doesn't inject variables by name, fwiw - you have to supply them to the @given(...) decorator.
We've also put a lot of work into type annotations, so you can get static checks for most simple problems 🙂
(with ParamSpec making more possible in 3.10)

[fiery arrow]

I think what they meant is that if you do

@given(x=text())
def test_foo(x: int):
    ...
``` a type checker won't be able to catch the mistake

I think ParamSpec won't solve this, unfortunately (unless you type text() to return a str). It doesn't let you "wrap" each parameter type in something

[hexed cloak]

You need to run mongodb when you run your tests

There's also stuff like mongita

[atomic zodiac]

There is no mongo instance connection from my CI build server 😦

mongita - I will check this out

thanks

[royal orbit]

Ah, yeah, that won't get caught 😔
Construction and composition of strategies is mostly statically checkable now, though, and there's a steady stream of useful peps (eg ParamSpec + Concatenate lets us precisely annotate the @st.composite decorator)

[fiery arrow]

Well, I don't think it's really an issue. The mistake would be caught as soon as you run the test, so it doesn't cost much

[royal orbit]

Yeah, it's not a big deal, but if we can support static checks (especially for editor squiggles) I do like to.

[pearl cliff]

indeed, fix error explained what i meant. but yes, that is the style i wish pytest would support!

[royal orbit]

saaaame, so many fixtures would be more ergonomic as context managers.

But Hyrum's Law is strict, we can't really change it now without breaking things for literally millions of people.

If I had a better solution I'd implement it, but with my pytest-core hat on incremental change (e.g. deprecating common misuses or confusions) is clearly the way to go.

🤞 we never end up with irrecoverable regrets about some core part of Hypothesis' design...

[pearl cliff]

context managers or hypothesis/click-style decorators-that-pass-kwargs

fwiw you could implement it as a new alternative api and keep supporting the old one

but that might be too much work

[royal orbit]

Best option IMO is to implement it as a downstream plugin which can be "blessed" once it's gone through trial-by-users for a while.
Too much work is the real problem, my backlog is something like a decade fulltime at this point. I might pitch some ideas at the PyCon sprints though...

[final fable]

Hi all,
I should write a test for the rest API with following properties:

• It has REST endpoint: /short/<ARG>.
• can be any combination of exactly two alphanumeric characters, e.g.:
  "Ab", "12", "a2".
• Returns json with 'uid' key and value of exactly 32 alphanumeric characters,
  e.g.: {"uid":"855f938d67b52b5a7eb124320a21a139"}.
  Would you please help me? I've not ever written test. how should I test it?
  Thanks

[gloomy cipher]

You basically need to write code that executes your code. That's what a test is. You pretend you are the consumer/user of your code, then you can run "assert" statements on the results. pip install pytest to get started.

[final fable]

thanks

[thick fjord]

yo, any links where i can learn about essentials of unittesting in flask ? Most vital stuff only

[gloomy cipher]

this looks promising @thick fjord https://stanford-code-the-change-guides.readthedocs.io/en/latest/guide_flask_unit_testing.html

and this: (i run this inside a private browser session so I can read it without paying for a subscription) https://medium.com/analytics-vidhya/how-to-test-flask-applications-aef12ae5181c

lastly https://testdriven.io/blog/flask-pytest/

[final fable]

Hello all,
I would appreciate it if you could introduce me a useful book regarding unit test(and generally test) in Python.
Thanks,

[maiden pawn]

it is written with examples in C#, but that's not really an issue in my opinion

[final fable]

Thank you. it looks great

[proper wind]

Hi I’m looking to learn about TDD. Any books you recommend? Thanks

[maiden pawn]

Technically original book for this goes from Kent Beck

But

Python has its own book

I am not sure which one is better 🤔

I went only through the part of the first book

356 reviews in amazon for Kent Beck
https://www.amazon.com/Test-Driven-Development-Kent-Beck-ebook/dp/B095SQ9WP4

60+100 reviews for python
https://www.amazon.com/Test-Driven-Development-Python-Harry-Percival/dp/1449364829
https://www.amazon.com/Test-Driven-Development-Python-Selenium-JavaScript-dp-1491958707/dp/1491958707/ref=dp_ob_title_bk

Python is surely fresh one, 2017 year

huh, in good reads, Kent Beck has even 3917 ratings
https://www.goodreads.com/book/show/387190.Test_Driven_Development

Kent Beck book is a bit old, goes to 2000-2002 year.

I should probably read them both and compare 🤔

[proper wind]

Awesome, just started reading the Testing Goat, but maybe I read the og book because I can’t go wrong with reading about tests, right?

[maiden pawn]

I liked very much reading Unit testing and best principles from Khorikov book

it allowed me to understand where I should concentrate my testing efforts

before reading the TDD

[proper wind]

Ok great, didn’t notice the book. The order of book reading: Unit testing -> The Testing Goat book -> TDD by Kent Beck

[maiden pawn]

it sounds like a good plan

[proper wind]

Thanks

[fading needle]

hello everyone! is there a way to test class based views in Django with pytest? could someone please provide me with any resources since i can not find any ( i am only seeing functional based views getting tested )?
Thanks!

[proper wind]

Hi I’m doing my first project. I think it’s difficult for me to do testing before writing code. Can you please forgive me for not following the Testing Goat?

[scenic hawk]

You must repent.

[pearl cliff]

the problem with unit testing is that most people test units that are too small. i bet you can come up with one reasonable invariant property of your application, right?

(practically yes of course just start and figure everything out as you go)

[proven elm]

The testing borg collective

[proper wind]

If I write code that works for similar to a calculator. I should write the tests first using pytest. I’m trying to look for basic information to do a basic test on a basic application. The books I’m reading to see intensive for a beginner like myself.

It is soo hard for me to understand the tests. Surely there’s a simple guide on how to get started for testing

[pearl cliff]

when you design your application, try to separate "functionality" from "user interface" as much as possible. that way you can write tests for the core calculator functionality without having to go through the entire user interface

it helps to think of your application as developing a library for yourself, and thenw rapping it

[proper wind]

Sure I can do that. I can write a separate file. What can I do after that?

[pearl cliff]

it doesn't have to be a separate file

it's just a way of thinking about your design

[pearl cliff]

the "core" of your application should be as much "pure functions" as possible, only doing user interface stuff at the "edges"

that way you can just write tests for those pure functions

[proper wind]

Awesome I will stick to doing for the script I’m writing. I can just use unitest for now. Apparently it’s not industry standard to use unitest though?

[pearl cliff]

either pytest or unittest is fine. i think unittest might be good to start with, because people do use it. and spending some time with unittest helps make pytest seem better by comparison!

however i think pytest really is better, despite there being a few things i don't totally like about it

fortunately pytest can run unittest test suites too

iirc it doesn't support a couple of the more sophisticated unittest features, but you aren't likely to use those anyway

[proper wind]

So what would be the best course of being a better tester. Use pytest?

Just learn this library well?

I appreciate the advice, I’m more confident now

[pearl cliff]

i suggest picking one and learning it

the other thing about unittest is that you don't typically need to worry about "plugins" like you do with pytest

i'd go with unittest to start

[proper wind]

Ok then transition to pytest after I get more comfortable

Thank you

[bitter wadi]

Hey @dusty torrent!

It looks like you tried to attach a Python file - please use a code-pasting service such as https://paste.pythondiscord.com

[thin robin]

@proper wind using the AAA or 3A test pattern is helpful (Arrange-Act-Assert). Here is an example in pytest:

# ./src/punctuation.py
def remove_punctuation(string: str):
    """Remove all punctuation from a sentence except for spaces"""
    return "".join(c for c in string if c.isalnum() or c == " ")

# ./tests/test_punctuation.py
def test_remove_punctuation():
    # arrange
    input_data = "Hi, this is 1 example."
    expected_output = "Hi this is 1 example"

    # act
    output = remove_punctuation(input_data)

    # assert
    assert output == expected_output

[fair elm]

Getting 0% coverage in sonarqube for python project which used nosetests to create coverage.xml file.

Can anyone help me with it , strongly stucked to it since a week.

What could be the reason for above mentioned query?

[weary quarry]

First thing would be to check sonar's output; is it consuming the .xml or raising a warning?
Next thing would be; Are the paths in the converage.xml file pointing to the code as defined in your sonar setup? (will sonar find the code from your project layer if it follows the paths). I had an issue with tox while testing an installed library that the paths pointed to tox's venv and not the project files.

[fair elm]

It's giving warn in sonar output , mentioning :

WARN: The report '/apps/dftjenkins/jenkins_node/workspace/v_AutomationAuthenticationScript/coverage.xml' seems to be empty, ignoring. '{}'
org.sonar.plugins.python.EmptyReportException: null

[weary quarry]

Is it an empty file?

[fair elm]

yes , it says it's an empty file

couldn't able to get it why .. It should be with coverage reports

[weary quarry]

I meant, if you look at that file is it empty?

[fair elm]

yes, It says it's empty, shall we connect here?

But nosetests should generate coverage.xml file .. if it tells all test as OK

[weary quarry]

It should be. I'm not super familiar with nosetest. Are you running it with the --with-xcoverage plugin or generating the coverage report directly with coverage xml?

[fair elm]

yes , running with --with-xcoverage plugin

that should generate coverage.xml file

[weary quarry]

I agree. I'm not sure why it wouldn't. nosetest isn't a tool I've used.

[fair elm]

Can you please elaborate it mpre

Next thing would be; Are the paths in the converage.xml file pointing to the code as defined in your sonar setup? (will sonar find the code from your project layer if it follows the paths).

setup.cfg is created for nosetests , and sonar-project is for sonar reference. I couldn't find any problem within it, could you plz check once!!

[nosetests]
verbose=3
with-xunit=1
xunit-file=nosetests.xml
with-xcoverage=1
cover-package=.
xcoverage-file=coverage.xml
exclude-dir=test

sonar.projectKey=PIN_API_hth
sonar.projectName=PIN_API_hth
sonar.projectVersion=1.0.0
sonar.language=py
sonar.sourceEncoding=UTF-8
sonar.python.coverage.reportPaths=coverage.xml
sonar.python.xunit.reportPaths=nosetests.xml
sonar.python.xunit.skipDetails=false
sonar.sources=backend/orchestrator, backend/worker, backend/worker/Relations.py
sonar.typescript.lcov.reportPaths=coverage/PIN_API/lcov.info
sonar.python.coveragePlugin=cobertura

This is jenkins stage code refering to coverage.xml for sonar

stage ('Test/Sonar') {
                    
                    steps {
                        // This specific examples expects that you have a Dockerfile.test that you build, then run to generate test results.
                        // Since different projects can vary however, you should make sure you use the solution that works best for you.
                        sh "docker build . -f Dockerfile.test -t python-test:${BUILD_ID}"

                        sh '''docker run -u 50001255:25 -v "$(pwd)":/app python-test:${BUILD_ID}'''
                        sh '''sed -i 's|filename="|filename="'"$(pwd)"'/|' coverage.xml'''

                        sonarScan('Sonar')
                    }

                    // Make test results visible in Jenkins UI if the install step completed successfully
                    post {
                        success {
                            junit testResults: 'nosetests.xml', allowEmptyResults: true
                        }
                    }
                }

A warning is raise from sonar side :

WARN: The report '/apps/dftjenkins/jenkins_node/workspace/v_AutomationAuthenticationScript/coverage.xml' seems to be empty, ignoring. '{}'
org.sonar.plugins.python.EmptyReportException: null

[weary quarry]

That part doesn't matter right now. You need to figure out why nosetest isn't generating a coverage report first.

[fair elm]

nosetests is working fine, I guess so It says below, but creating empty coverage.xml

docker run -u 50001255:25 -v /apps/dftjenkins/jenkins_node/workspace/v_AutomationAuthenticationScript:/app python-test:31
..............................
Ran 30 tests in 4.618s

OK

[fair elm]

Yes, but I'm clueless at this moment.. I've given 4 days time to investigate, even I tested few of my thoughts.

[weary quarry]

cover-package=. does the package directory have a __init__.py in it? I'm clueless as well.

[fair elm]

No Preocts.. cover-package=. is considering entire application file/folder. and top of that there is no '__init__.py'. file

[fair elm]

Please let me know , if you get something to resolve this 🙂

[proper wind]

do you use kvm?

[fading needle]

Hey, has anyone experienced this response didnt redirect as expected: response code was 200 (expected 200)

[bitter wadi]

:incoming_envelope: :ok_hand: applied mute to @fallen moss until <t:1645462800:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

:incoming_envelope: :ok_hand: applied mute to @fallen moss until <t:1645463658:f> (9 minutes and 59 seconds) (reason: duplicates rule: sent 4 duplicated messages in 10s).

[viral pebble]

Hello,I have a question about tests. I have just created a Python module that allows me to manage docker containers but I have no idea how to write the tests.should I create mocks to simulate the responses of the docker commands or should I let the containers be created during the tests? (this will involve running the tests on a machine with docker)

[scarlet marlin]

It'd probably be most repeatable to set the tests up as docker-in-docker

there's some challenge to this but you should only have to solve it one time

https://devopscube.com/run-docker-in-docker/

you probably shouldn't be setting up mocks to mimic responses of the Docker containers, because then you'll be liable to just end up in a situation where you're testing your mock implementation, not actually testing whether your code does what you think it does to Docker containers

[viral pebble]

Ok, thank you very much. I am going to see that 🙂

[worthy flare]

anyone familiar with web automated testing using python & unittest? i need a bit of help, ty

[fair elm]

I couldn't able to integrate nosetests coverage.xml with sonarqube scanner. sonar gives status of 0% coverage .
Can anyone help me with it?

Situation is similiar to :

https://stackoverflow.com/questions/48741685/python-sonarqube-integration-with-nosetests-and-coverage-not-shown-cover-code

[kind meadow]

Is there a nice way with pytest to use the same parameters twice, one with the original value and one with a function applied to it? I know I can just modify the list of parameters beforehand, but I am wondering if there is like some magic way to just add a second decorator.

[potent quest]

could you send your current code?

specifically the decorators and signature