unit-testing

[jaunty magnet]

;compile

import time
import os

# os.system("cls")
inp = [
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬛⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬛⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬛⬛⬛⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
    "⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜⬜",
]


def nextgen(inp):
    nxt = []
    alive = "⬛"
    dead = "⬜"
    for n, m in enumerate(inp):
        inp[n] = inp[n].center(len(inp[n]) + 2, dead)
    inp.insert(0, dead * len(inp[0]))
    inp.append(dead * len(inp[0]))
    for y, i in enumerate(inp):
        temp = ""
        for x, j in enumerate(i):
            count = 0
            tmp=[]
            if y == 0 or y == len(inp)-1 or x == 0 or x == len(inp[y])-1: continue
            for n in range(-1,2):
                for m in range(-1,2):
                    if n==0 and m==0:
                        continue
                    tmp.append(inp[y+n][x+m])
            count = tmp.count(alive)

            if inp[y][x] == alive:
                if 2 <= count <= 3:
                    temp += alive
                else:
                    temp += dead
            else:
                if count == 3:
                    temp += alive
                else:
                    temp += dead
            if y == 0 or y == len(inp)-1 or x == 0 or x == len(inp[y])-1: continue                   
        nxt.append(temp)
    nxt=nxt[1:-1]
    for n in nxt:
        print(n)
    print()
    return nxt


nxt = nextgen(inp)
while True:
    nxt=nextgen(inp)

[maiden pawn]

Oh my.

That is master piece i will save for interviews about refactoring

[novel canopy]

that's a lotta squares

[maiden pawn]

https://github.com/darklab8/darklab_examples/blob/master/python/help/help_to_zhabiboss/before.py
I also liked this code piece

perfect for refactoring interviews too

the above ons is better though, for more hardcore. way more nestiness

[novel canopy]

lol

[tacit dawn]

hey guys i need help

can i post my code here

and ask for the problem ?

the code is about reversing the words on its place

hello my name to olleh ym eman

s=input("Enter the sentence")
l1= s.split()
z= len(l1)
l2=[]
a=0

for i in range(0,z):
k=l1[i]
l = len(k)
for a in range(-1,-l-1,-1):
h=k[a]

l2.append(h)

print(" ".join(l2))

[vale vault]

Building tests are somewhat kinda confusing I can't lie

[maiden pawn]

haven't had that feeling. writing tests is always kind of obvious.
hard things can be regarding them only creating good code architecture that will be easy testable.

if test fails, it is always great instead of problem. u discovered a problem your test was written for 😁

[hybrid ravine]

I have some problem with Django.
when i run the command
python manage.py test

I am getting errors like this on the image

But when i delete init.py from application directory the error disappears
I will only add that the project works, the file structure is fine and all applications are added to settings.py

[languid sentinel]

How to approach testing code that is accessing azure infrastructure? I'm using Python SDK to access files on Storage Account

[maiden pawn]

well..Clean Architecturing acess to Azure components into some class, that u will mock in the rest of your code for unit tests.

Leaving turnable integration tests capable to check functionality of this class wrapper

in this way u will have tests running in majority without need for Azure access, and even covering functionality of this Azure interacting code too

[weak knoll]

Hi folks,
I am writing python for the past 6 years, but I never writen any tests
I would like to learn how to convert my code skills to be more TDD instead of YoLo 😄

Any good resources, advise where to start, etc ?

[maiden pawn]

Read it first to get understanding why and for what

[maiden pawn]

Then this for more practical head on

and then continue learning OOP/Design patterns/Code architecture. with books like Head First Design Patterns, Code Complete by McConnel, Clean Architecture by Robert Martin and etc, in order to get code architectural part of writing your code in order to have it testable

[weak knoll]

Thanks a lot

[viscid basalt]

Is it possible to provide a default value for fixture in pytest if it's not available? 🤔

# worker_id fixture might not be available here
def my_fixture(worker_id: str = ...) -> None:
   ...

[dusty hamlet]

Fununit 2.1.0 has released! Fununit is a functional, declarative unit-testing library for testing pure functions in Python.

• functional: fununit has a functional API, which means it has great composability and flexibility
• declarative: fununit lets you describe unit tests in a fully declarative manner
• cuts out all code duplication: fununit removes up to 100% of unnecessary code duplication between tests

if you want to write unit tests that look like this:

multiply_tests = tests(
    ["Math"], "multiply", multiply, [
        ("zero", (0, 6), 0),
        ("identity", (1, 2), 2),
        ("positive", (3, 4), 12)
    ])

then check out fununit on github (https://github.com/johnmpost/fununit) or pypi!

[marble wagon]

I have some tests in a tests/ dir, and when I'm on linux I can run PYTHONPATH="$(pwd)" pytest so that my tests can access the modules in my root dir - What is the windows equivalent of PYTHONPATH="$(pwd)"?

[maiden pawn]

use Git Bash(comes with Git CLI) as terminal https://git-scm.com/downloads , or https://taskfile.dev
both allow to have linux shell scripts running at windows

don't learn Windows. 😆

[marble wagon]

ah okay thanks

Yeah i dislike windows

[placid heron]

Is there a way to use TestCase / LiveServerTestCase server in Selenium to interact with the page?
Selenium runs the real server webpages now but i want to use the test env

[royal orbit]

Generate tests programatically for property-based testing.
Are you thinking of Hypothesis integration? Not sure how you'd provide an idiomatic API, specifically how properties would be described when you can't just list the expected output value.

[ember maple]

anyone aware of a library for making in-memory pathlib compatible paths
i want to get away from always writing to temporary folder when testing interaction with specific fs state

[proud nebula]

https://pypi.org/project/pyfakefs/ ?

[ember maple]

thats monkeypatching all the things instead of providing path objects for something in memory

[proud nebula]

ah

[ember maple]

i'd like to have soemething like a DictFs({"foo":{"test.md": "# content"}}) an the use it like pathlib

its a major blunder that pathlib has no concept of "filesystem"

[proud nebula]

That sounds like a pretty great lib idea :P

[ember maple]

well, no can do from me ^^ im stretched already ^^

[proud nebula]

Know that feeling. I'm writing a totally new i18n system for Django now.. for some damn reason heh

[ember maple]

oh thats damned

@proud nebula i18n apis are close up with date apis ^^

[proud nebula]

It's not actually that horrible heh

Maybe I overstated the case a bit. I'm still going to use the Django gettext/gettext_lazy functions. Just a much better string collection system, with no dependencies to unix gettext (and the ENORMOUS hacks Django has to enable that)

And just in general vastly improve the DX of the thing. Cut the number of gettext() calls in models to almost nothing by default for example.

[proud nebula]

I do take on big idiotic projects a lot it seems. Mutmut and hammett are kinda crazy if you think about it.

hammett is very neglected though

[ember maple]

well, i wanted to start cogs of testing in 2020 to externalize gnarly part of pytest for reuse, its 2023 now an im about to go on parental leave with the 2nd kid instead

[proud nebula]

Hah, well there's no big idiotic project quite like a kid eh? Kinda puts the other things in perspective

[ember maple]

well, unlike software, with a kid there is a good chance that with good maintenance the idiocy resolves in less than 30 years

[dusty hamlet]

I think it's doable. You're right though, it would have to be slightly different than the basic fununit notation. But I think you could just list the expected output value - you'd just be generating a bunch of separate unit tests with different parameters and expected outputs. I'd need to do some more brainstorming to see if this works for a variety of property-based testing scenarios. I might try to build this in as a feature if I come up with something

Hmm, it definitely depends on what kind of property you are testing... I'm gonna have to think more about this

[viscid basalt]

Do you plan on adding python 3.11 support to mutmut? 🤔

[proud nebula]

Yes. But the mutmut db is based on pony which breaks on 3.11 so I have to wait for that.

[proud nebula]

https://github.com/ponyorm/pony/pull/671

[viscid basalt]

I see, I'm wondering why did you decide to use an orm for mutmut?

[proud nebula]

Faster dev time.

[viscid basalt]

True, is pony orm just a personal preference? Never used it myself

[proud nebula]

It looked pretty nice and I thought going for the django orm was a bit much :)

[viscid basalt]

I personally would prefer sqlalchemy over django orm 🤔

[proud nebula]

I wouldn't:) It's more complex and annoying for everything basically :)

[viscid basalt]

Pony orm querying style looks a bit weird, IMO, not that it's a negative thing
Queries and models in sqlalchemy 2.0 are now fully typed, which is a big plus

[uneven beacon]

i could have sworn this was working a few weeks ago, but for some reason my pytest suite stopped being able to import the parent tests/ package: ```py
..\project> pytest tests\cogs\names
===== test session starts =====
platform win32 -- Python 3.11.1, pytest-7.2.1, pluggy-1.0.0
benchmark: 4.0.0 (defaults: timer=time.perf_counter disable_gc=False min_rounds=5 min_time=0.000005 max_time=1.0 calibration_precision=10 warmup=False warmup_iterations=100000)
rootdir: ..\project, configfile: pyproject.toml
plugins: asyncio-0.20.3, benchmark-4.0.0, cov-4.0.0
asyncio: mode=Mode.STRICT
collected 24 items / 1 error

===== ERRORS =====
_____ ERROR collecting tests/cogs/names/test_cog.py _____
ImportError while importing test module '..\project\tests\cogs\names\test_cog.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
tests\cogs\names\test_cog.py:7: in <module>
from ... import *
E ModuleNotFoundError: No module named 'tests'
===== short test summary info =====
ERROR tests/cogs/names/test_cog.py
!!!!! Interrupted: 1 error during !!!!!
===== 1 error in 0.16s =====file structure is setup as:py
project/
src/package/...
tests/
cogs/names/
test_cog.py
from ... import *
init.py # contains fixtures
pyproject.toml
... # poetry backend
[tool.pytest.ini_options]
addopts = "--import-mode=importlib"``` i also tried creating a fresh environment with virtualenv + poetry install and also downgrading pytest to my last used version, 7.2.0, to no avail

[viscid basalt]

@uneven beacon Why would you from ... import *? 🤔

[uneven beacon]

i have my common fixtures in the tests root

[viscid basalt]

You can use conftest 🤔

[uneven beacon]

hmm ill look into that, i just dont get why it was perfectly fine before until i took a break and it broke down

does conftest actually allow you to use variables defined inside it, or is it just to resolve the parameter names? my init.py exports some constants that id like to use for typehinting the fixtures correctly

[viscid basalt]

It allows you to use fixtures

[uneven beacon]

so not enough to cover my desires... 😔

[viscid basalt]

Maybe it does, what kind of variables do you need?

[uneven beacon]

just the fixture return annotations, in my case one type alias and one class from the source package

perhaps worth noting that python doesnt crash if i import the test files directly, e.g. py -m tests.cogs.names.test_cog

[viscid basalt]

I'd avoid using relative imports, unless module is on the same level, e.g. from . import some_module

You should put your fixtures in conftest.py, so they're automatically discovered by pytest

[proud nebula]

You normally don't import fixtures

[uneven beacon]

apparently moving my fixtures into conftest.py while retaining a single __init__.py in the tests/ root fixed it, but for convenience im still importing conftest directly so i can still have my fixtures' typehints

[viscid basalt]

🤔 By having their typehints you mean?

Honestly mistyping fixtures in your test might be an issue that's currently not solved, unless there's a flake8/mypy/pytest plugin for that:

@pytest.fixture
def some_fixture() -> int:
    return 42

def test_something(some_fixture: str) -> None:
    ...

[uneven beacon]

inferring it would be ideal but eh :shrug:

[viscid basalt]

There's a plugin for type checking 🤔
https://github.com/untitaker/pytest-fixture-typecheck

Well, it's like 5 lines and could be ported to any codebase

[uneven beacon]

it seems like the plugin wouldnt cover much that your type checker cant do

1. if you correctly typehinted your fixture but misused it, either your type checker complains or your test raises a runtime error
2. if you gave a generic typehint like Any or didnt typehint the fixture, typeguard wouldnt complain and you'd get a runtime error if the fixture was misused
3. if you mis-typehinted your fixture, either your type checker complains first, typeguard raises a runtime error slightly earlier than your test would have, or it catches a typehint bug that doesnt impact your test

[viscid basalt]

2. Forbid any 😃

Typecheckers are not aware that fixtures are somehow related to your test parameters

Also your fixtures may come from different modules and/or plugins, I would be too lazy to import them

[proud nebula]

"are not"? that seems rather broad. PyCharms built in type checker absolutely understands pytest fixtures

[viscid basalt]

It also removes type annotations when I rename them 😅

Which is "great"

I think it's a part of pycharm integration with pytest, but I don't think mypy/pyright support that, at least not mypy 🤔

[proud nebula]

You mean when you use the rename tool, it removes the explicit type declaration from your code?

[viscid basalt]

Yep 👀

It removes annotations from tests where that fixture is used:

@pytest.fixture
def number() -> int:
    return 42

def test_number(
    number: int,  # type hint would be removed
) -> None:
    pass

[proud nebula]

Weird. Have you reported that?

I don't understand how you can stand having that much type annotations everywhere though heh. Might as well write Rust.

[viscid basalt]

No, it's already reported:
https://youtrack.jetbrains.com/issue/PY-44112/Renaming-a-pytest-fixture-deletes-typing-information

Pycharm still doesn't infer context manager types:
https://youtrack.jetbrains.com/issue/PY-36444
https://youtrack.jetbrains.com/issue/PY-29891

It's been years

[proud nebula]

Well.. there are really a ton of issues it's been years for :/

They broke del unused a few years ago for example. Quite annoying since it used to work.

That's my biggest beef with pycharm honestly, that they don't seem to have regression tests for stuff.

[lament heath]

Oooh that's amazing.

[lament heath]

I also hate that they have a very loose grip on the community. I.e. mypy is made by python community and pyright is developed by significant members of python community and is officially supported by typeshed.

As a result, pycharm is pretty outdated in many regards.

[lavish vault]

Is it pretty common to do test driven development. I cant seem to get in the rythm of it and habits of not adding tests just keeps lurking

[maiden pawn]

Read test driven development by Kent Beck
That book teaches feeling of how often tests are needed being written

[maiden pawn]

Ergh... I am pushed to write tests by habits:

1. without tests I have no fast way to confirm this code works
2. without tests hard to localise errors after refactor or new feature addition
3. I need writing tests to write new feature, because I explore syntax of new stuff in test

[twin tusk]

@lavish vault David Beazley wrote a great article on testing a few years back that's worth a read: https://www.usenix.org/system/files/login/articles/login_fall17_10_beazley.pdf

[thorn bluff]

oh, that was interesting, thanks

[twin tusk]

For me, the major advantage of TDD is, even if you don't follow it fully and don't write tests first, it puts the question "how would I test this?" permanently in your mind as you code.

[maiden pawn]

Exactly!!!!!1111111oneoneome 😁

You can skip TDD, but TDD will not skip you

[final nest]

anyone with experience using locust?

[proud nebula]

I updated my old article on accidental primary key alignment with postsgres code and a pytest+django implementation. Thought it might save someone else some headache: https://kodare.net/2018/08/28/intermittent-tests-aligned-primary-keys.html (I got bitten by this badly yesterday)

[sleek sluice]

Hello!

im new to unit-teting.

I am looking to test a Flask-Rest-API. What would be a good practice to test out all the endpoints and class methods;

Would it be better to run a mock-up version of the server, and test out the endpoints using various HTTP requests?

or should i focus on testing out the return values of each function/method of each endpoint?

[ember maple]

strongly depends on the intended logic behind it - i'd recommend to move tricky stuff into function that you can call directly for a test , and then having a few select test that ensure everything works together

for the tricky bits one really wants the surrounding shenanigans out of the way, for the rest ensuring all things work together is a bit important, also make use of type annotations to have your back

[sleek sluice]

Great thank you!

If i may add another question; is there a convention on how to deal with objects that are initiated and imported from other modules?

Like a logging instance i imported into the specific module i want to test.

It would throw an error, because the defined relative directory does not exist from the current called module

[ember maple]

It's not clear what you are doing there

Either use absolute imports or figure the relative one

A pastebin of the folder structure may help

[maiden pawn]

Python unittest in Mock patch can mock even global stuff if desired
But it is a bad practice

Good practice is writing Class with inclusions of objects via Composition and overriding necessary stuff during test setup

Also logging library is not supposed to break during testing. U have wrong module import if it happens

[sleek sluice]

I think my issue is that i cant seem to override modules that are irrelevant for the specific unittesting but are necessary for general Runtime of the application.

here is a clarification in code:

the file/module i want to test a function in

import logging, re
from logging.handlers import TimedRotatingFileHandler


def get_logger():
    logger = logging.getLogger(__name__)
    logger.setLevel(logging.DEBUG)
    handler = TimedRotatingFileHandler("./logs/historie.log", when="midnight", interval=1)
    logger.addHandler(handler)
    
    return logger


log = get_logger()


def return_one():
    return 1

and the unittest file:

import unittest
from unittest.mock import patch


class TestExt(unittest.TestCase):

    @patch("ext.get_logger")
    def test_return_one(self):
        import ext
        self.assertEqual(ext.return_one(), 1)


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

this unittest still throws an error because it is still calling the get_logger() function.

Basically i want the ext.py module to be imported but the setup of the logging object to be ignored. And patch didnt help in this case

EDIT:
i just learned through research that the code inside of modules have to be run anyways. And there is only the option to avoid running code in a module is by mocking the whole module, but it will give no access to the objects/classes/variables... etc in the module.

It seems that i'll have to implement the if __name__ == "__main__": conditional in most of my modules to avoid such errors.

[dense bough]

My understand of how this works:

• patch assigns to ext.get_logger, making it available to the test;
• import ext rebinds ext;
• At this point, if get_logger is not defined in the module, then ext.get_logger is undefined.

[maiden pawn]

to patch get_logger, patch needs to patch "module_name.get_logger"
in any case it is wrong approach, because affecting global state is not really good way to go for testing

[hollow basin]

@thick star have you tried using pytest with pytest-mock?

your code

def start_time():
    current_time = datetime.now()
    return current_time
    
class TestStartTime(unittest.TestCase):
    def test_start_time(self):

        expected_time = datetime.datetime(2023, 2, 8, 10, 0, 0)
        datetime.now = MagicMock(return_value=expected_time)

        result = start_time()

        self.assertEqual(result, expected_time)

with pytest and pytest-mock

def test_start_time(mocker):
    mocker.patch("datetime.datetime.now", return_value=datetime.datetime(2023, 2, 8, 10, 0, 0))
    expected_time = datetime.datetime(2023, 2, 8, 10, 0, 0)
    result = start_time()
    assert result == expected_time

pytest is better than unittest in pretty much every way. a lot of people treat it as the default unit testing framework, but it's easier to develop it separately from the language itself.

[proud nebula]

Patching datetime is best done with time-machine also

[hollow basin]

idk what that is 😛

[maiden pawn]

i remember briefly it before

frozen time or something like saw somewhere

freezegun

@pytest.mark.freeze_time('2019-10-04')

[maiden pawn]

https://github.com/spulec/freezegun

from freezegun import freeze_time

freezer = freeze_time("2012-01-14 12:00:01")
freezer.start()
assert datetime.datetime.now() == datetime.datetime(2012, 1, 14, 12, 0, 1)
freezer.stop()

controlling flow of time 😆

@freeze_time("Jan 14th, 2020", auto_tick_seconds=15)

even speed

[proud nebula]

don't use freezegun

It's pretty much dead, has lots of bugs, and the basic approach isn't really good

(I'm funnily enough one of the maintainers.. but I've given up)

https://github.com/adamchainz/time-machine

[maiden pawn]

😁

[dawn frost]

Have any of you tried JetBrains Aqua? I keep seeing it, but don’t really understand how it’s different from normal IDEs yet

[dawn frost]

ig not 😦

[weary raptor]

maybe ask in #editors-ides

[thick star]

@hollow basin Very belated thanks for you reply. Sadly though, pytest complains in oh so many words about the same problem: TypeError: can't set attributes of built-in/extension type 'datetime.datetime'

[plucky ice]

Hey is this for selenium? Channel

[proud nebula]

If you want to mock time, use time-machine https://github.com/adamchainz/time-machine

[proud nebula]

This is a pretty random channel honestly. So you can try at least.

[dawn matrix]

Hey, I am testing a Class where I read some json data from a file. When running the test, the path for the files is wrong (because my tests are one stage lower than the class). How can I solve this problem. I am using pycharm.

[proud nebula]

because my tests are one stage lower than the class

that made no sense

Maybe show a small code example?

[proper wind]

hey guys

im making code for linux on windows

how would i test it?

[proud nebula]

You mean that should work on both?

[plucky ice]

I am trying to do selenium set speed 10

But it isn’t working in python

I don’t get it been trying for hours

Keeps saying invalid syntax

[plucky ice]

Thanks

I am just trying to slow it down because it’s so fast

[maiden pawn]

@proud nebula i can guess he means, path files are working incorrectly at linux, because he is having miss matches in capital/lower case file names, which linux notices due to being case sensetive, and windows not

answer is: run tests in Github Actions / Gitlab CI 🙂 they will rerun quickly for LInux. Or develop already from linux / or work from case sensetive mounted filesystem

[maiden pawn]

same answer as above: run tests in Github Actions / Gitlab CI 🙂 they will rerun quickly for LInux. Or develop already from linux

[proud nebula]

So you can't write valid python. Show us the error! And the code.

[proud nebula]

That's a pretty long guess

[proud nebula]

You shouldn't guess wildly imo. That's a bad habit and not one to be displayed in public :P

[dawn matrix]

Sorry, I badly expressed. I won´t deploy to Gitlab or working with linux. This is a project for my University ( so developing on windows :D) . I meant that I have the following folder structure: poker(folder) here is my python script and my json files. In the poker folder I have a nother folder for tests. And here are my scripts. Now if i run the scripts, the paths for the json files are not found. The paths are (./json_file_1.json). I need to change the paths to (../json_file_1.json). Now the tests would work, but if I only run the script, it wont find the paths.

[mellow thicket]

actual unittest file will know where it is .. construct relative path from the unittest file's parent directory and convert to absolute and a) it will work b) it will show you where your error is if you are always using relative paths ?

[proud nebula]

That doesn't sound related to linux/windows at all?

[maiden pawn]

that is why i pinged question owner :b to ensure he would correct if i guessed wrongly

[maiden pawn]

use pathlib library and write files in some fashion that it not changable with you moving across folders
choices

1. for example you can write path like... str(pathlib.Path(__file__) / "json_file_1.json")

that will make it searched strictly relative to the file from where you have written this script

tests will not be able to change such path

2. also in a same way you can discover project folder once for all your scripts pathlib.Path(__file__)
   and having other scripts using this once discovered path to build based on it their paths

from project.settings import project_folder
path = str(project_folder / "folder" / "file.json"

[cobalt ore]

is there a way to test conftest.py ? I can run a particular test with pytest -k '<filter>', but was wondering if there was a way to just run the conftest with pytest 🤔

note - i dont' mean write a separate test for conftest, just to run pytest without actually running any tests (which to me means it'd run the conftest to setup and not actually do anything). But i might be mixing some things up

[proud nebula]

I don't see the point. Many things in there won't be called unless a teat is run anyway.

[cobalt ore]

I think you're right, I thought it was an ok idea but in hindsight it wasn't at all😅

[craggy jolt]

where do i use coverage run -m unittest discover

and what is the general process for test harnesses in python?

what does that mean in the context of python

is it just unittest?

[proud nebula]

Personally I would say: use pytest. It's much nicer.

[rotund dagger]

I have an object which is either a string or a pytest.mark.structures.ParameterSet. How can I import ParameterSet so I can do an isinstance(obj, ParameterSet) ?

Variables:
m: MarkDecorator [pub]
| .args: tuple (2)
| .kwargs: dict (0)
| .mark: Mark [pub]
| | .args: tuple (2) [pub]
| | | 0: 'meow'
| | | 1: list (1) [pub]
| | | | 0: 'kitty' <---------------
| | .kwargs: dict (0)
| | .name: 'parametrize'
| .markname: 'parametrize'
| .name: 'parametrize'
m2: MarkDecorator [pub]
| .args: tuple (2) [pub]
| | 0: 'meow'
| | 1: list (1)
| .kwargs: dict (0)
| .mark: Mark [pub]
| | .args: tuple (2) [pub]
| | | 0: 'meow'
| | | 1: list (1) [pub]
| | | | 0: ParameterSet [pub] <-----
| | | | | 0: tuple (1) [pub]
| | | | | | 0: list (1) [pub]
| | | | | | | 0: 'kitten'
| | | | | 1: tuple (0)
| | | | | 2: 'P'
| | .kwargs: dict (0)

from

#!/usr/bin/env python3

import pytest

m = pytest.mark.parametrize("meow", ['kitty'])
print(m)

m2= pytest.mark.parametrize("meow", [
    pytest.param(['kitten'], id = 'P')
])
print(m2)

print(type(m2.mark.args[1][0])) results in <class '_pytest.mark.structures.ParameterSet'>

however, isinstance(m2.mark.args[1][0], pytest.mark.structures.ParameterSet) results in

AttributeError: 'MarkDecorator' object has no attribute 'ParameterSet'

[hearty oriole]

Hello

Anyone who is good at mocking a database insertion for a unit test? Because I'm having difficulties with mocking one 😅

[proud nebula]

The first rule of mocking is to not do mocking :P Have you tried really really hard to not mock?

[fiery arrow]

Suppose I'm writing API tests for a parental control service. I might want to create a "family" like this:

sample_family = {
  "id": "family-smiths",
  "name": "Smiths",
  "parents": [
    {
      "id": "dad-smith", ...
    },
    {
      "id": "mom-smith",
      "authorize_devices": [
        {
          "id": "mom-smith-iphone", ...
        },
      ],
    },
  ],
  "children": [
    {
      "id": "child-alice-smith", ...
    },
    ...
  ],
}

The problem being, of course, that I can't hardcode the IDs, they're generated randomly. Is there a neat way of defining such things so that I'm able to e.g. pass them to pytest.mark.parameterize?

Like, in the test data I might include some more information on how to create the family, then the parents, then the children etc. -- that's where I'd get the IDs, step by step. And in the end I want to check that the result looks like the above^ (more complicated in reality of course)

[maiden pawn]

i would advice inspiring with code approach of factoryboy (each sub objects is created automatically with sub factory. all attributes are created with fake data, you are having precision to create fake objects only in relevant parameters you need. It can help to minimize data.
https://factoryboy.readthedocs.io/en/stable/introduction.html

And you are certainly can be using pydantic https://docs.pydantic.dev/ library as main source of your... reusable body requests.
you can define your objects once as its BaseModel, and with obj.json() or obj.dict(), it will make nested serialization to your end desired view
Together with utilizing Pydantic Field(default_factory=list | lambda: write your own generator) you can essentially create all your request bodies

[plucky ice]

Hey is there software testing channel for appium

?

[fiery arrow]

hmm, not sure how pydantic will help here

I don't want to create random data in the test. The IDs will be generated on the server side

[ember maple]

its exposed as pytest.ParameterSet these days

we probably should ensure it reports pytest a s module to ease things for people

[rotund dagger]

My colleague suggested import _pytest.mark.structures.ParameterSet, and it worked, but it feels so dirty

[spring nebula]

anyone have any ideas on how to make pyautogui not go for the same image twice when both images are the same images

[ember maple]

sometimes thats a valid suggestion, as long as its accompanied by a upstream ticket to ensure a importable name if not existing

[jolly tangle]

Hello super quick question I am writing code in pytesting and in the tutorial they have 2 folders unit and functional. What goes in each folder and what is the definition of each? I am using a flask web application.

[proud nebula]

The exact definition varies by project/person. But mostly unit is small purely functional tests. While "functional" is often end to end. Yea, the names suck :)

[jolly tangle]

@proud nebula By "end to end" I assume you mean like testing routes. By "functional tests" I assume you mean functions. Is this correct?
Thanks

[proud nebula]

No, unfortunately not.
"end-to-end" is the same as "functional tests"
"unit tests" are for simple functions that don't touch the database

Like I said, the names are quite bad.

End-to-end usually means testing the entire stack, by running a web browser to test functionality. You want as little as possible of those because they are super slow.

[viscid basalt]

How do you actually name tests for code that interacts with database? 🤔
e.g.

def some_function(connection) -> int:
    return connection.execute("select 1")

[earnest sequoia]

test_query_<query_name>?

[viscid basalt]

I mean unit, integration/etc

[proud nebula]

test_some_function?

using pytest obviously

[viscid basalt]

I meant what you would "classify" these tests as

Is it a unit test? Integration test?

[proud nebula]

oh, "integration" is what we called them at a previous job

You have to have a pretty big test suite before you split these up and care much about the names though imo.

[viscid basalt]

I could always use marker if I need to, currently my tests directory structure is the same as my main code

[viscid basalt]

Would it make sense in case of a function/class like this to move query itself into a separate entity? 🤔

[earnest sequoia]

if you are only testing a specific query, I would say it is a unit test of the query itself

if it involves multiple queries or a whole transaction then I think it would be an integrated test

[viscid basalt]

To be honest when creating web applications it could be hard to not use a database in a lot of tests

[earnest sequoia]

yeah I'm doing the same thing, the ORM makes it harder to separate database from the business logic

however

using data transfer objects between modules can greatly reduce the need for creating database objects as the data transfer objects themselves can be easily used as stubs

[viscid basalt]

And using something like repository feels like just creating another abstraction layer on top of orm 😵‍💫

[viscid basalt]

I use dtos as inputs into the application core/business layer from various adapters (for example from web api), but I return ORM objects directly from it

Unless it can't be represented as existing model, then I make a dto for a result of that specific action

[vivid cosmos]

is testing in flask necessary

[viscid basalt]

"Is testing necessary?" - It depends

[vivid cosmos]

like I just dont see any point of me testing in flask because of the nature of it

im doing it as a side project to get a internship

[viscid basalt]

@ember maple Is _pytest.* considered a public api? 🤔

[ember maple]

nope

[viscid basalt]

Why not?

[ember maple]

see the underscore?

[viscid basalt]

@pytest.fixture(scope="session", params=["asyncio", "trio"])
def anyio_backend(request: SubRequest) -> str:
    return request.param

How would I annotate request then? 🤔

I also use this hook/plugin to check if test parameters are annotated properly:

def pytest_runtest_makereport(item: Function, call: CallInfo[Any]) -> None:
    if call.when != "call":
        return

    for fixture_name, type_ in item.obj.__annotations__.items():
        if fixture_name not in item.funcargs:
            continue
        typeguard.check_type(fixture_name, item.funcargs[fixture_name], type_)

Function and CallInfo are a part of private API too

https://github.com/untitaker/pytest-fixture-typecheck

[ember maple]

hmm, ineed, its missing in https://github.com/pytest-dev/pytest/blob/main/src/pytest/__init__.py#L24 - in that case, import from the internals and lets see if there already is a issue for exposing it

both function and callinfo are exported in pytest already

ah, https://github.com/pytest-dev/pytest/issues/8763 is the related issue

[viscid basalt]

Function, CallInfo are available though, but pycharm still suggest importing them from private api 😃

[ember maple]

seems like subrequest is not the way to go atm

hmm, oh, thats a issue that may need collaboration from pycharm,

[viscid basalt]

There's also another very annoying bug, related to pytest - renaming any fixture erases type annotations from tests where its used

[ember maple]

ok, sub-request is not exposed for good reasons it seems (aka even fi you used it, the types woudl be wrong

[viscid basalt]

I see, by the way - is there a plan to add an api similar to pytest.mark.parametrize but that could be used programmatically? 🤔

Similar to pytest.fixture(params=...)

[ember maple]

pytest_generate_tests ?

the parameterize marker is implemented by that hook, fixture params too

[viscid basalt]

I see, I should probably take a look 👀
Some of my parametrized data also depends on other fixtures/configuration, so having a function return a list of parameters would be better

[ember maple]

parametrize cant use fixtures

it happens before setupstate is made, setupstate is where fixtures get set up and torn down

[viscid basalt]

I know, that's why I'm saying that I need something different

[viscid basalt]

Why you don't want to test a flask application?

[vivid cosmos]

idk what to test

[viscid basalt]

You can test your internals, or responses from your flask app, If you're using templates then testing responses is probably not worth it

But you still can test internals of your application

[ember maple]

@viscid basalt a plugin class that uses pytest_configure/unconfigure + provides the additional hooks may work, then you contain the state in your plugin object

[vivid cosmos]

so i just do a brief tests

?

[viscid basalt]

Brief?

[vivid cosmos]

brief overview

not like very very comprehensive

[viscid basalt]

Are you making an API or does it use templates?

[vivid cosmos]

templates

[viscid basalt]

You can test that your templates at least render properly
Also it shoud be easier to test that your templates actually receive correct data

If you move logic away from flask endpoints you should be able to test it separately

But if your site is not dynamic (e.g. it doesn't use any kind of storage/database) then there's not much to test 🤔

[vivid cosmos]

its dynamic

[viscid basalt]

Then test it 😉

[maiden pawn]

is good quality of code necessary for smth?
not really for simple very small stuff that will be thrown away and never improved for new features, or ever taken care by other developers. Otherwise better to take more care

[ember maple]

hmm, testing tends to feel unnecessary until the lack of tests bites you in the back cause our house was made of cards instead of something more solid and you never knew ^^

conversely adding tests typically brings lots of little hidden human errors to the light, just try to break it a little, and oh boy it will - in particular in the face of change

[maiden pawn]

And typing... Feels unnecessary in the same way

And once u get hooked to typing, u can't live already without it

[ember maple]

yeah, refactoring is way better without the blind duck

[jolly tangle]

@proud nebula thanks

[gloomy crow]

When running a bunch of tests with pytest, is there any way to automatically continiue on failed ones?

[earnest sequoia]

Doesn't it already do that by default? You have to pass -x flag for pytest to terminate on the first failing test

[maiden pawn]

it does it by default

[gloomy crow]

Hmmm, i guess it could be --pdb flag then

[ember maple]

Is there any maxfail in the afdoptd?

[gloomy crow]

I think not, cant check untill tomorrow morning(in12h or so :D)

What happens is that it enters debug in each failed test, so i have to type c

Which makes me unable to trully go afk while i run all tests(3000+) 😄

[hearty oriole]

I have already solved this issue 🙂 No I intended to mock because I am only interested what the actual code expects, hence I am totally not interested in the DB. Therefore, I mock the interaction with the database.

[proud nebula]

5 days to respond. Still missed the point ;) You can change your code to have pure functions and teat those. Again: avoid mocking.

[proud nebula]

There's --lf which runs the tests that failed last time. That's probably what you want.

[hearty oriole]

Apologies for my late response. Regardsless if I make them into pure functions they are still dependent to the database, therefore I only mock the database object.

[brazen needle]

A pure function depends only on its input. Your unit test can provide that input. In practice, that input would be retrieved using a database query, but your unit test does not have to do things that way. If you structure your code correctly, your unit test will not have to mock the database interaction. This is usually a good idea because mocking is difficult.

[weary quarry]

At some point you will have a function who has the dutiful job to handle a crud operation to the database. That function's behavior should also be tested. Depending on the setup I've simply replaced the database connection with an in-memory sqlite3 connection or have a database spun up in a docker to the side. I "mock" the database by replacing the production database with something testable.

Both options I've used are made simpler by having said database connection be an injected dependency to the function/class that uses it. That makes replacing that connection much easier in the test. The function doesn't care about what connection it runs the crud on, only that it has one.

And at that point you could just replace the connection with a MagicMock() and assert for calls, but it's almost as easy to spin up a testing database. You have the bonus of testing behavior and function of your crud operations with the latter. :)

"Just don't mock" is a neat idea. A lot like "don't use if statements". It works, until it doesn't.

[brazen needle]

I like the idea of using an in-memory sqlite3 database. I feel like mocking tends to be fragile; it's hard to get the mock's behavior right. But in a way, database interaction is something that you can't actually "unit test" in the strict sense because it's fundamentally about an interaction between your code and someone else's.

[weary quarry]

I agree. The unit test should be testing the behavior of the function. If I give you X and Y I assert you perform Z. I just see it as an added bonus that you can also assert your queries are doing what they should in the case of the database.

Change the third party target to, say, an API and I change my tune. Mock the call, assert the behavior of your code, and don't try to test the API. :)

[proud nebula]

that's not a great function, you want at least SOME sleep() in there!

[viscid basalt]

I think you could wait for N seconds and raise a specific Timeout error if it didn't complete 🤔

This can be done using threads or signals on linux

Or tasks if you're using asyncio 🤔

I think it would also be a good idea to add timeout prameter to wait_until_not_busy

[ember maple]

timeouts are needed, threading primitives should be used ^^

if you arent allowed to change the function, then i hope its part of a bad exercise and not something in production

[jolly tangle]

I am running some pytesting code with the command in the terminal by typing python - m pytest. How do I print, print("something") from conftest.py fixtures to the terminal in Visual Studio Code? My setup is similar to this https://gitlab.com/patkennedy79/flask_user_management_example . Here is the link where I got the uploaded code https://testdriven.io/blog/flask-pytest/ . Though my code is a decent amount different.

Side note every folder has a __init__.py file. I just didn't include it in the diagram below.

testing folder 
|-----conftest.py
|
|                                         
|                           
|------------------------------ functional   folder    |
|                                                      | 
|                                                      | -----------------------------------test_routes.py
|                                           
|                                           
|---------------------------------------- unit folder  |-----------------------------------test_functional.py
|                                                      |-----------------------------------test_models.py

Thanks

[craggy jolt]

whats the structure of professional testing in python usually? is it just a file with different functions that test their respective functions?

[weary quarry]

Sure, usually. I don't know about the whole "professional" part of it. Given main.py I'll have a test/main_test.py and in that test file I'll have as many tests as needed to cover each unit and any branches.

I prefer main_test.py over test_main.py simply because it makes autocomplete easier.

[rare atlas]

having main_test.py also is nicer for sorting filenames if they're in the same directory

[topaz socket]

How do I print()
pytest --capture=no

[lethal river]

https://gist.github.com/hasandiwan/61d5c7cc81eaa8ed9b6ead35f2451357 any idea?

[weak drum]

boy test_sentiment.py doesn't even have proper indentation

Also, what's f'/sentiment?url={url}' do you have a local sever running for /sentiment? if not, that's not a valid url

[jolly tangle]

Thank you very much. How did you know to do that?

[topaz socket]

I got same problem and simply searched the web...

[jolly tangle]

@topaz socket how did u phrase the search?

[weary quarry]

"display print output pytest" brings up several articles. The pytest documentation has the cli option as well. -s is the equivaent to --capture=no.

[vapid field]

idk if this the right place to ask but i'm new in coding and i wrote a simple code to choose a random letters and i want the output to be in insta like to check if the username is valid or not but idk how to connect them anyone can help?

[daring coral]

Hello all. I'm struggling wrapping my head around pytest's fixtures and the class scope. I'm testing a client/server project. While testing the client I need to spin up an instance of a mock server, this is done by calling an exe. I want the mock server to persist for all of the tests in a given class and terminate after all client tests are run. Would this be done with pytest.mark.usefixture before the class definition or do I place the mock server fixture in each of the test methods inside the class. thanks in advance for any help.

[maiden pawn]

you can use @rare sentinel.fixture(scope=session, autouse=True)

someone named themselves pytest :\

[daring coral]

Thanks for the reply. I will give that a try in the morning.

[sharp sparrow]

guys, how can i monkeypatch a function in my own module for pytest?
code structure

src
 - module-name
    - utils.py
      - func1(): ...
tests
  - conftest.py
    - replacement_func(): ...
    - conftest_func_name(): ...
  - test_demo.py
    - test_demo1(conftest_func_name):

so, basically, i need to monkeypatch the func1() function from my module with replacement_func() function in the conftest.py file
and i have no idea how to do it
any ideas?

[proud nebula]

Step one of monkeypatching for tests: avoid

[ember maple]

Monkeypatch can't replace the function in other places where it was imported

It's recommended to avoid monkeypatch if feasible

[sharp sparrow]

then how do i have to achieve this thing?

[split field]

do the callers of func1 do from module-name.utils import func1, or do they do from module-name import utils and then call utils.func1()?

[sharp sparrow]

its a module, it can be done in anyway
but I'm actually doing something like this in the conftest.py

import module_name

# inside a fixture function 
monkeypatch.setattr(module_name.utils, 'func1', replacement_func)

and its not working

[split field]

if the callers of func1 do from module_name.utils import func1, then monkeypatching module_name.utils.func1 after they've done that isn't enough - you need to also monkeypatch the_module_that_already_imported_utils.func1

[sharp sparrow]

how? can u show me how it would look like?

[split field]

monkeypatch.setattr(module_that_did_the_import, 'func1', replacement_func)

https://docs.python.org/3/library/unittest.mock.html#where-to-patch probably explains this better than I can

(that's written for unittest.mock.patch, but everything it says about where to patch applies equally to the monkeypatch fixture)

[sharp sparrow]

u mean like, this thing would do the monkeypatching for scenarios like this?

module_name.func1()

i mean, i do import stuff in __init__.py but i don't import that function
so that the user only has to use the function like this.module_name.utils.func()

im confused

[split field]

if the user does from yourmodule import func1, then they have a reference to func1. If, after that point, you change yourmodule.func1 to be a different object, that doesn't affect the reference that user has to the original func1

[sharp sparrow]

i need to patch that function only for a test

[split field]

it doesn't matter why you need to patch it. What matters is that you need to replace every reference to it (or at least, all of the references to it that might be called by the test)

[sharp sparrow]

@split field then do i need to change it manually in the test function like this? (what u r telling is too much complex for me)

import module_name

def replacement_func(): ...

def test_demo1():
  module_name.utils.func1 = replacement_func

  # do stuff

 # and i dont know what to do to  get it reverted back to its original function 

[split field]

I told you to call monkeypatch, and what you're showing doesn't call monkeypatch. I don't know how you got to that from what I said.

!e ```py
import math
from math import pi

print(f"{math.pi=}")
print(f"{pi=}")

Now monkeypatch math.pi to a new value!

math.pi = 3

print(f"{math.pi=}")
print(f"{pi=}")

[bitter wadi]

@split field :white_check_mark: Your 3.11 eval job has completed with return code 0.

001 | math.pi=3.141592653589793
002 | pi=3.141592653589793
003 | math.pi=3
004 | pi=3.141592653589793

[split field]

if you've done from math import pi, then later monkeypatching math.pi doesn't affect the already-imported pi

[sharp sparrow]

sry man, im confused
so, im only using that function only in one test file like i've mentioned here (this is in conftest.py)

import module_name

# inside a fixture function 
monkeypatch.setattr(module_name.utils, 'func1', replacement_func)

and as far as i've seen, i need to change every reference of the func1 right?

and also, im not directly using that function by the way
I'm actually calling it by calling another function inside the same module (sorry for telling this late)

# module_name.utils
def func1(): ...
def func2():
  func1()

# test_demo.py
def test_demo1.py
  func2()

# conftest.py
import module_name

# inside a fixture function 
monkeypatch.setattr(module_name.utils, 'func1', replacement_func)

i hope that i've given enough details to understand this thing even more clearly

[split field]

can you share your actual code?

your simplifications aren't helping people understand, because I think you're dropping relevant information.

[sharp sparrow]

I'm not having my laptop right now, so let me just tell you even more

the func1 has an infinite while loop in it like this

module_name.utils
def get_command_func():
  return sys.stdin.readline

def func1():
  get_command = get_command_func()
  while True:
    command = get_command()
    if command == 'exit':
      return
    # doing some command execution

def func2():
  # some code in here
  func1()

and all i need to do is to mock the user input for the tests

for that, i am trying to do something like this

# conftest.py
import module_name, pytest

a = iter(('help', 'exit'))

def replacement_func():
  yield next(a)

@pytest.fixture
def sample_input(monkeypatch):
  monkeypatch.setattr(
    module_name.utils, 'func1', replacement_func
  )

so, is there any way that makes this code work with some sample inputs?

[split field]

so every call to func1 is made by func2, and func1 and func2 are in the same file?

[sharp sparrow]

s

i mean, yes, every call to func1 is made by func2, and func1 and func2 are in the same file. and i'm using func1 by calling func2

[split field]

then absolutely none of the advice that people have been giving you about where to patch was actually relevant, and we've all been misunderstanding what your problem is

[sharp sparrow]

so what's the actual problem then? mocking stdin?

[split field]

I don't know - what's the problem?

what you're doing looks like it could work to me.

I'd rewrite ```py
a = iter(('help', 'exit'))

def replacement_func():
yield next(a)
aspy
def replacement_func():
yield from ("help", "exit")

but other than that, what you're describing looks OK to me.

[sharp sparrow]

it just looks ok, but that is not working at all
that's the thing

[split field]

does your test use the sample_input fixture?

[sharp sparrow]

yes, it does

[split field]

then I'm sure I won't be able to help more without actually seeing the real code.

[sharp sparrow]

the test function signature is this -

# test_demo.py
def test_demo1(sample_input, capfd):

[split field]

it sounds like you're patching the right place, the patch that you're describing should work, so if it isn't working, the reason why is likely something that's in your real code that isn't in your simplified examples.

[split field]

It could be related to this if multiple test functions want to use that fixture. With how you've written it, the iterator would get completely consumed by the first test that uses it, and then every other test wouldn't get any input

[sharp sparrow]

im not having my laptop rn, so i can't post the code btw

but i can post it here about 12 hours later (im in clg)

[split field]

ok. Without seeing it, the best I can say is that what you're describing sounds like it ought to work. If the only caller of func1 is in the same module as func1, and you're patching func1 in that module, then every call to it should call the thing you've patched in, instead.

[sharp sparrow]

im using that fixture only once among the whole test cases

@split field ok, thankfully i've the code in github

this is the func1 that im talking about - https://github.com/GhostOps77/click-repl/blob/GhostOps77-patch-1/src/click_repl/utils.py#L344

and this is the func2 - https://github.com/GhostOps77/click-repl/blob/GhostOps77-patch-1/src/click_repl/utils.py#L353

and the test function would be something like this (i do need to capture stdout, so im using capfd)

from click_repl import repl
import pytest

def test_demo(sample_input, capfd):
  # here, i have some basic click commands implemented here, something like this - https://github.com/GhostOps77/click-repl/blob/GhostOps77-patch-1/tests/test_repl.py#L7-L24

and i haven't pushed the conftest.py code, but this is the exact code that i've implemented

import click_repl, pytest

a = iter(('help', 'exit'))

def replacement_func():
  yield next(a)

@pytest.fixture
def sample_input(monkeypatch):
  monkeypatch.setattr(
    click_repl.utils, '_get_command_func', replacement_func
  )

[split field]

_get_command_func is supposed to accept two arguments, and your replacement_func doesn't. Based on what you're showing, I would expect the patch to be applied successfully, and I'd expect a TypeError exception on this line - https://github.com/GhostOps77/click-repl/blob/GhostOps77-patch-1/src/click_repl/utils.py#L392 - when the patched in replacement_func gets called.

[bitter wadi]

src/click_repl/utils.py line 392

get_command = _get_command_func(isatty, session)```

[sharp sparrow]

yeah, that's why i haven't implemented it in the actual code yet, i just have it in my system

so maybe, this would work

def replacement_func(a, b);
  # same thing

and now tell me, what should make the mocking stdin work?

[split field]

you can patch your own object in as sys.stdin if you want to - like an io.TextIO object

[sharp sparrow]

does it really work in my get_command func? cuz as u can see in the code, im also using prompt session's prompt function. so does it also work with the sys.stdin?

[sharp sparrow]

and also, another question, what if i want the stdin related test to be parameterized? i mean, i would like to pass in specific std inputs, and expect some specific output from it, how can i pass the args to the fixture?

[split field]

I don't know - I've never used prompt_toolkit, so you'd need to check the code. If it does an import sys and then uses sys.stdin, then it would work.

[split field]

It would probably be easiest to use a regular function instead of a fixture, at that point. It's possible for a fixture to use parametrizations, but that's gonna be harder for people to read than keeping things simple: make a function that that takes monkeypatch and your input, and then uses monkeypatch to patch stdin or your _get_command_func function to return that input.

def set_stdin(monkeypatch, user_input):
    monkeypatch(sys, "stdin", io.TextIO(user_input))

@pytest.mark.parametrize("user_input", ["help\exit\n", "foo\nbar\nbaz\n"])
def test_something(user_input, monkeypatch, capfd):
    set_stdin(monkeypatch, user_input)
    repl(...)

[sharp sparrow]

can i define the set_stdin function in conftest.py so that it can be accessed in all other test files?

[split field]

I think it would be better to put it in a util.py file in the same directory as conftest.py, and do from .util import set_stdin wherever you need it

in theory you can import stuff from conftest.py too, but that's not what it's for, and I'd be worried about stuff running twice that shouldn't

[sharp sparrow]

in my opinion, set_stdin should not be accessible by the user. so i would better define that in conftest.py

[split field]

"the user" of what?

it should be in your tests directory, and that whole directory is for maintainers, not for users

[sharp sparrow]

it should be in the test directory
i dont think you've got what im saying
the set_stdin function should be only a patch function for testing, it should not be published to public
but i hope that i can define in conftest.py
and it doesn't matter cuz im gonna run the test for std input only once among all the other tests

but still, i'll check ur code after like, 8 hrs

[split field]

I do get what you're saying. I think you didn't get what I'm saying: I would put it in a file called tests/utils.py instead of a file called tests/conftest.py

[mortal galleon]

hi
I am writing a unit test for a function that uses the request library. Ive tried chatGPT and youtube but it just wont work... It seems that the mock values doesnt get passed to the function. This is my loggs:

2023-03-03 06:52:47,214 ERROR Type: <class 'Exception'>
2023-03-03 06:52:47,214 ERROR Details: <MagicMock name='get().raise_for_status()' id='139916181325952'>

When i print response, from:

response = requests.get(url,parameters)                #Issue the GET request
print(response)

I get:

<MagicMock name='get()' id='139916181232800'>

Error:

AssertionError: None != [{'code_review_id': 123, 'title': 'Test code review'}]

first time doing this, and im new to python + ive rewritten this class now 100 times. so it might look stupid but here it is:

class TestCrawler(unittest.TestCase):
    @patch('test3.requests.get')
    def test_makeRequest_success(self,mock_requests):
        mock_response = Mock()
        mock_requests.status_code = 200
        mock_response.text = ')]}\'\n{"data": [{"code_review_id": 123, "title": "Test code review"}]}'
        mock_requests.get.return_value = mock_response
        mock_response.raise_for_status.side_effect = Exception('Not found')
        data = makeRequest(url, qParam)
        check = [{'code_review_id': 123, 'title': 'Test code review'}]
        self.assertEqual(data,  check)

same if i print status_code:
<MagicMock name='get().status_code' id='139875748252544'>

[mortal galleon]

used responses library instead

[ember maple]

you never configured mock_requests, the first thing you did after taking it as parameter is to replace it with something entirely else that wont do the work

[ember maple]

also i strongly recommend to never use requests_mock - its a footgun ^^

[sharp sparrow]

i've tried this, and its not working at all

[proud nebula]

Don't use those dirty words.

[kind meadow]

Can you elaborate on what can go wrong? I've only used it once in a small project.

[jolly tangle]

I am pytesting some code in python flask and getting an error sqlalchemy.exc.IntegrityError: (sqlite3.IntegrityError) UNIQUE constraint failed: user.username Can someone help?

Since this this pytesting it is not as obvious solution. Any help is appreciated

Here is the error

https://pastes.io/iijimkme1k

Here is the code

def test_register_page_get(client, new_user, False_output_check_if_user_already_registered):
    
   response = client.get('/register', follow_redirects=True)
    assert response.status_code == 200
    assert b'register' in response.data

    with app.test_request_context(): 
       number_of_users = User.filter_by(username=new_user.username).count()    
       if number_of_users > 1:
               db.session.rollback()        
 
     
        try:   
                db.session.add(new_user)
                 db.session.commit()
            #False_output_check_if_user_already_registered(new_user)
        except:
            db.sesion.rollback()
        else:
            db.session.delete(new_user)
            db.session.commit()

I even tried changing to

 from sqlalchemy import exc
if exc.IntegrityError:
          db.session.rollback()     

instead of

number_of_users = User.filter_by(username=new_user.username).count()    
if number_of_users > 1:
      db.session.rollback()        

in the example above and I am still getting the same error.

[ember maple]

The mock surface tends to be wrong when replacing requests in that manner

Hynek wrote a couple of good blog posts on it

See https://hynek.me/articles/what-to-mock-in-5-mins/

[ember maple]

Sorry didn't mark the reply correctly

[potent quest]

(haven't looked at the link yet) is it specific to requests_mock or just all http mocks, eg aioresponses for aiohttp and whatnot?

[ember maple]

It depends, I'm not familiar with all the apis, if they resemble requests mock then it applies

The blog post gives good examples and further reference, i strongly recommend digesting it first

[potent quest]

will do

[shrewd oak]

@split field Couldn't you inject fake data in stdin instead of setting up a mocked stdin stream?

[sharp sparrow]

so, do u know how to inject data through stdin?

[ember maple]

How is the stdio of prompt toolkit actually being set up?

https://python-prompt-toolkit.readthedocs.io/en/master/pages/advanced_topics/unit_testing.html looks completely covered

[shrewd oak]

Stdin.write?

Just like any other IO stream.

[shrewd oak]

Sorry, I had some thoughts about the stdin mocking problem. I think, it would be better to mock readline using unittest.mock.patch instead of monkey patching stdin. Just encapsulate the unit test body into an with patch("sys.stdin.readline", side_effect=[str1, str2, ..., "exit"]):. You can easily combine unittest mocking capabilities with pytest.

[sharp sparrow]

im sry man, but do u know how to use those fixtures?
can u help me on that?

[ember maple]

Never used them, but the examples seem complete enough to use them

[sharp sparrow]

where did u get those examples? i only see the fixture code in there

[ember maple]

The firs code example shows how to use the objects, the fixture example shows what part to put into a fixture,

[sharp sparrow]

yeah, then how do i send input through the input pipe then?

[ember maple]

Based on the fixture in the doc, the fxture provides the api to send in input, and the first code example shows how to send input to that

[sharp sparrow]

i mean, i can't call a fixture, right?

how do i get the input stream object from a fixture?

[ember maple]

Fixtures get passed into functions accepting them

Gimme a sec to type that in on the phone

def test_input(mock_input) :
  mock_input.send_text(...)

[sharp sparrow]

ok, now what about the output?

capfd?

[ember maple]

That may need some research in how to use the output parameter of prompt toolkit, which I cannot do this weekly

[sharp sparrow]

so, as far as i've seen, this prompt_toolkit's pytest fixture is only useful to test only the prompt toolkit stuff

and i've no idea how to send the prompt toolkit objects inside the app_sesion to run them all

[sharp sparrow]

after sending a value to that generator, no change has happened
definitely a thorough research should be done

[shrewd oak]

It has already been said: Avoid monkey patching. Remove the get_command_function, the get_command and the replacement_function, use sys.stdin.readline directly and then use @patch or with patch(...):to patch readline.

[proud nebula]

"avoid monkey pathing [...] monkey patch"?

[shrewd oak]

@proud nebula Monkey pathing???? Were did you read that?

Step one of monkeypatching for tests: avoid

[proud nebula]

and then use @patch or with patch(...):to patch readline
That's monkey patching

[shrewd oak]

Grmpf. Strictly yes, but the API behind is a little bit more elaborate than what you can with the manual approach.

[proud nebula]

Well you should avoid that too imo.

[shrewd oak]

OK, thats leading to a personal discussion I can't continue today. However, I will look into that 5 min video and maybe my brain gets enlighted after that. Anyway, interesting topic.

[weary quarry]

Are you willing to expand on why your flat answers appear to be "do not mock"?

I'm curious to your reasons here. In the majority of experience I've had mocking is an invaluable tool. The only downfall I've found is that some tend to be looser with their test code than their implementations, causing issues. Smells in those cases are, generally, caught and corrected in review.

[proper wind]

here's a very simple question, if it doesn't belong here let me know. I'm testing a really big database-making program for my Object-Oriented Programming class (which is in python for some reason?) and it takes forever to put the first dozen or so inputs into pycharm over and over

can i automatically program a specific set of user inputs to start it off so i can test the later things without having to write the same stuff over and over?

[maiden pawn]

Use pytest for testing

[autumn bronze]

If you're taking stuff through input() and can't be bothered setting up pytest, you can just put the text on different lines in a file and pipe it into python

cat file_with_inputs.txt | python3 script.py

On Linux like systems, and you can probably find a windows equivalent

[maiden pawn]

Or u could replace taking input from input, with inserting input through generators or smth during tests. Can be more convenient during testing

[proper wind]

honestly i haven't been using python for longer than 1.5 months so this is a HUGE help, i'll check it out!!

[autumn bronze]

Sure, like I said this is to do it without changing much in the script

There's obviously better architectural changes that can be made

[maiden pawn]

Technically there is pyexpect to interact with input(), but replacing during tests input to smth else is better.

Python unittest.patch with mock can directly hack input() to return sequence of certain values in a row

That would be better solution

[maiden pawn]

Pytest + mock library for the win
https://docs.python.org/3/library/unittest.mock.html

[autumn bronze]

I don't disagree, I was just providing a "get it working" approach

[proper wind]

oh this is like 500 lines of code and i am new to python

is pytest gonna need like a ton of rewriting

[maiden pawn]

Ergo... Inputting from file is not clean. Mocks are cleaner more precise surgical instrument

[maiden pawn]

Pytest is backward compatible with Inbuilt unittest lib (that is assuming u complain about rewriting unittest tests)

[proper wind]

can i like set a given list of inputs that it'll take and it'll just run through the next one for each input()?

[maiden pawn]

With unittest.mock? Yes

[proper wind]

ok cool

it's not like an application, it's a program in terminal?

ok yeah

[maiden pawn]

Feel free use default unittest instead of pytest.. If u work without libraries

[proper wind]

yeah i'm not getting that advanced anytime soon

oh if i can just use unittest yeah

ok this seems like much more in-depth testing stuff for individual variables: is there a way i could just implement this in a normal run of the code?

[maiden pawn]

Certainly. Have an easy hack...

def input_generator():
  yield "ABC"
  yield "123"

iterator = iter(input_generator())

def input():
  return next(iterator)

This code will replace Inbuilt input with input returned in sequence from input generator

Technically... If would be shorter and better done in unittest.mock

But i am at phone, and can't remember from the top of a head how to do it with mock 😁

[proper wind]

ok, thank u so much!

wait what if i just want it to do it for a certain number of inputs and then after that i put them in

[shrewd oak]

No need for generators, if you pass a list to side_effect. Example: input_generator = MagicMock(side_effect=["abc", "123"]).

There is this discussion about using (monkey)patch or not. However, none provided a non-patch alternative. I'm curious, too.

[proud nebula]

Avoid. Not "never ever use". But dependency injection is better.

[fickle solar]

I have 3,000 test cases. Each one is a page of text. Do people typically use SQLite to store these?

[shrewd oak]

My test cases are usually within a source tree.

[proud nebula]

What do you mean by "test case"?

because what you said doesn't make sense with the definition I'd use

[fickle solar]

A page of text from a document. The expected result from a page is a list of tuples.

[fickle solar]

As standalone files? Could do that. Annoying for git though, and the file system might not support more than 20,000 files in a folder. We're expecting to have several million test cases eventually

[shrewd oak]

@fickle solar I think we're talking about different dimensions.

... And for me a test case is a bunch of code not a bunch of test data.

Which means the bunch of code is there to test something specific and orthogonal about something under test.

[brazen needle]

I might call these "test inputs" rather than "test cases." You have enough of these that you need to think about how to organize them. A database is probably a good idea.

If you're really planning to have several million of these (but how are you planning to make them? This part doesn't make sense to me), then I doubt you want to put them all into your source tree. But you're going to want to keep them under version control. You could consider having two separate source trees, one with all your code, and one that simply holds test data. You could still arrange your test data as individual files as long as you can think of a hierarchy of some sort (even if it's just numbering: Test inputs 0 to 999 go in 0, test cases 1000 to 1999 go in 1000, etc.).

If you take that approach, you will need to be very careful about how you set up testing to make sure that you run all the tests you're supposed to be running.

[proud nebula]

Sounds iffy to me. Wouldn't you be better off with a more focused testing system?

[fickle solar]

Maybe. It's a machine learning algorithm that extracts information from text.

We need to make sure it meets a threshold accuracy on the dataset we have, so that we can report it to clients.

It's not all machine learning, and when we update the algorithm we need test inputs to see if there's regression.

[brazen needle]

It sounds to me like you need two different testing regimes, then. You need conventional testing to make sure that your code functions the way it's supposed to (doesn't crash, handles invalid inputs gracefully, etc.). But you also need algorithmic test cases to ensure that your algorithm functions correctly, and those are what you're asking about. Is that right?

And I'm guessing that you have some method for generating test cases and known correct output, and that's why you can plausibly have millions of test cases.

[proud nebula]

Oh. Ugh. Yea. Ok.

[fickle solar]

That makes sense yeah. Thanks guys.

[supple flume]

I've heard the tear unit testing a lot but what does it actually mean

[maiden pawn]

This book explains greatly what it is.

[maiden pawn]

and this one shows more head on approach

[supple flume]

is it a complicated topic

[maiden pawn]

TLDR summarization: Unit testing is a way to increase quality of your code
Making better code architecture and easier refactoring it to cleaner state.
important in order to have maintainable code

[supple flume]

why is called unit testing

[maiden pawn]

more formal definition: Unit tests are way to test a small unit of your code

ensuring its expected working according to written test specifications

[supple flume]

Wait you can do ?

with an IDE or with code

[maiden pawn]

both 🙂

but well, with code first

[supple flume]

that seems helpful

[maiden pawn]

you can run tests from console, or from IDE too 🙂

and having visual debugging interactive process during their run if you want (when running from IDE)

[supple flume]

thank you for the explonation

unit testing seems very useful

[maiden pawn]

just pressence of unit testing = jumps your code quality by three times 🙂

if you think how you are going to test your code before you write working code itself => Quality is doubled and tripled in addition

[supple flume]

so I have to write the tests myself

or import a module

[maiden pawn]

u need to write tests yourself

[supple flume]

got it thx

[maiden pawn]

while importing module for unit testing 🙂

unittest in python is default... but usually if you install any third party modules, better testing with pytest

[brazen needle]

pytest is just better than unittest. It's easier to use and has more features. I wouldn't recommend unittest to anyone at this point.

[maiden pawn]

it makes sense to learn... for small scripts you know, which do not have third party libs installed

pytest is backward compatible with all unittest stuff anyway

unittest.mock is splendidly awesome, and makes sense to use with pytest too, instead of pytest obfusactions for this

[weak drum]

using robot framework, but unable to get any output although i set the logfile to sys.stdout; verified this works when calling on the class it self ```python
import pexpect
import sys
from robot.api.deco import keyword

class main:
def init(self):
child = pexpect.spawn("/bin/bash", encoding="utf-8")
child.logfile = sys.stdout
self.child = child

@keyword    
def ls_cmd(self): 
    self.child.sendline("ls -la")
    self.child.expect("#")
    
@keyword    
def cleanup(self):
    self.child.close()

[shrewd oak]

Totally agree. I also use pytest for testing and unittest.mock for mocking together.

[solemn flint]

Hi, can some one help to find understanding in mock functions?

@patch('report_generate.build_report',  return_value="some_data")
def test_init(mock_build_report):
    assert build_report('path_to_filese_for_report_building') == "some_data"
    mock_build_report.assert_called_once()

perhaps there is some description besides the official documentation, for easier understanding

[proud nebula]

You forgot to explain what you think is confusing

[solemn flint]

I want to understand the principle of work in general

[proud nebula]

That's... not really actionable. Imo it's better to first get to a point where you can ask a coherent question.

[tacit crown]

Python3 unittest import hell

Have scoured the usual sources and can't find a solution that works that doesn't smell.

My project directory structure:

project/
    src/
        a.py
        b.py
    tests/
        tests_1.py
        tests_2.py

File a.py imports b:
import b

tests_1.py and tests_2.py need to import a and b:

import unittest

from src import a, b

What's the "best practice" way to get this to work without smelly hacks?

[uneven beacon]

shouldnt that just work? a gets imported which imports b, then b gets imported which returns the previously cached module to your test scripts

[proud nebula]

Those imports are super wrong. You can't import the same module through two different paths "a" and "src.a". That's very very bad.

[tacit crown]

Then how?

[tacit crown]

It doesn't "just work".
Either imports into the tests works but complains that "b" can't be imported,
or,
imports into a.py works but no tests are found.

This stackoverflow question covers the same topic, https://stackoverflow.com/questions/1896918/running-unittest-with-typical-test-directory-structure , but as the later comments to the accepted answer say:

This won't work if antigravity/antigravity.py imports from a module under the same antigravity directory.

[proud nebula]

pick one. Presumably not the one beginning with "src", because that's not what you want.

[proud nebula]

from src import a, b should be import a, b

You are confused about how the python path works. It's not like directories at all.

[tacit crown]

That gives me: ModuleNotFoundError: No module named b

[proud nebula]

Yea, so your project isn't correctly set up then. I personally really dislike the "src" directory thing. It makes everything complex and confusing for no reason.

I would move a and b up a level and delete the src dir

Then in the same directory as a, b, tests folders, you can just type pytest and everything will work by default

Otherwise read this: https://stackoverflow.com/questions/50155464/using-pytest-with-a-src-layer

[tacit crown]

You mean like this?

project/
    a.py
    b.py
    c.py
    ....py
    readme.md
    LICENSE
    .pylintrc
    project.ui
    .gitignore
    tests/
        test_1.py
        test_2.py
        test_n.py
    docs/
        build/
        source/
        Makefile

Isn't it rather messy to have the project's python files mixed up with so much other stuff? I don't see many Python projects doing this other than very small projects.

[tacit crown]

Recently, pytest has added a new core plugin that supports sys.path modifications via the pythonpath configuration value.
but I'm using unittest, not pytest. Does unittest support sys.path via a configuration value? I don't see that in the unittest documentation.

I do appreciate your help @proud nebula but I've been going round this ballpark for 2 days and I'm astounded that there are so many recommendations to put the project's .py files into one folder, tests.py files into another, but no definitive way to make that work with unittest.

[proud nebula]

https://github.com/django/django <- django does this.

[proud nebula]

I mean.. I wouldn't use unittest anyway. Pytest is better pretty much in every way. So... you're kinda shooting yourself in the foot and then being angry at others for your foot hurting. Just stop shooting yourself.

[tacit crown]

Not angry 🙂 but maybe a bit frustrated.
Unittest is part of the standard library, but you're saying it's not fit for purpose?

[proud nebula]

https://github.com/benoitc/gunicorn <- also does this
https://github.com/Textualize/rich <- this too
https://github.com/davidhalter/parso <- and this

I don't get at all where you're getting this from that "big" python projects use a src folder. I just looked at my top 4 most known dependencies and all of them do what I said.

[proud nebula]

Yes. Use pytest.

Also, use a good project structure :P

You want to have just one dir with your projects stuff in it in the top level folder anyway. You shouldn't have a.py and b.py, you want to namespace your stuff like your_project/a.py

Only exception is if your entire project is literally one .py file. Like polib has that for example.

[tacit crown]

I'm not really sure what you're suggesting.
Do you mean I should do this:

project\
    myproject\
        a.py
        b.py
    tests\
        test_1.py
        test_2.py

[proud nebula]

yes. Then from myproject.a import some_symbol

This is idiomatic python.

[tacit crown]

from myproject.a import some_symbol in a.py, b.py and test_n.py?

[proud nebula]

It will be the same syntax in all files yes

I only use import x type imports in some VERY rare cases. It's better imo to stick to from x import y style imports always

[proud nebula]

yea, don't forget the __init__.py file

in both tests and myproject dirs

[tacit crown]

That looks encouraging. I'll try that, but it'll take a while to test as I'll need to change a lot of files.
Do you happen to know where __init__.py is documented for Python3 (I'm getting confused by all of the Python 2 references when searching).

[proud nebula]

With pycharm refactoring tools it should be very fast

[proud nebula]

it's the same from like python 1 I'm pretty sure, so shouldn't be any different

[solemn flint]

https://docs.python.org/3/tutorial/modules.html

[proud nebula]

if you have a __init__.py in a dir, it means it's a python module. That's about it. Not much more to know.

[tacit crown]

I've found the search term Namespace packages - this looks promising.

[solemn flint]

https://stackoverflow.com/questions/448271/what-is-init-py-for

[proud nebula]

Just read what we posted here. It's enough.

[tacit crown]

Regular packages are traditional packages as they existed in Python 3.2 and earlier.

Do "regular packages" exist in Python 3.9 and later?

[uneven beacon]

my mistake on this, i didnt really think about your mixture of local/installed imports that deeply

[tacit crown]

That's not working for me.
I'm getting the error:
ModuleNotFoundError: No module named myproject
I've added __init__.py to all of the relevant folders.

[uneven beacon]

i think this layout is fine if your workflow is always installing your project in editable mode before doing any work

i.e. py src/ a.py import b b.py tests/ test_1.py import a pyproject.toml sh /myproject $ pip install --editable . /myproject $ pytest

[tacit crown]

but I'm not wanting to install it yet as the project is far from complete. I'm still developing the app.

[uneven beacon]

the point of an editable install is to let you work on the project (without reinstalling / tweaking pythonpath) while allowing your imports to be written in the same way they should be when installed

[tacit crown]

Thanks for the suggestion, but it seems to contradict the suggestions from @proud nebula as, if I understood him correctly, implied that just fixing the directory names and changing
import a
to
from mypackage import a
is all that is required.

I'd really like the canonical solution as it would clearly help many others that have the same question.

[uneven beacon]

this is known as a flat-layout, and the one you previously had (and im showing) is an src-layout
https://setuptools.pypa.io/en/latest/userguide/package_discovery.html
see also: https://blog.ionelmc.ro/2014/05/25/python-packaging/#the-structure
and: https://blog.ionelmc.ro/2017/09/25/rehashing-the-src-layout/

[proud nebula]

And your current working directory is....?

[proud nebula]

He doesn't know any of that stuff. Keep it simple!

[proud nebula]

just ignore that please

[tacit crown]

I've tried several. How should I be launching the app ("a.py") so that it finds
from my_package import b

[uneven beacon]

i kind of already demonstrated it, you dont need to know much about the details to enjoy its utility

[proud nebula]

Well.. you can't just first not give the full error message, then reply like that "several". There's no way I can answer a question like that.

If the current working directory is the dir that contains myproject and tests, then running pytest will just do the right thing

[proud nebula]

there's no utility to that stuff that isn't simpler and better without the src folder

simple is better than complex

[uneven beacon]

and both are still simple layouts

[proud nebula]

one requires modifications to the python search path. The other does not.

One approach is generally useful to understand how python works, the other is not.

[proud nebula]

you didn't say what you did to get that error even

[uneven beacon]

well the links i provided already provide arguments for and against its usage in projects

[proud nebula]

Seems like a long list of why not to me 🤣

[tacit crown]

Sorry, but you've lost me.
Say this is my project:

project\
    my_project\
        a.py
        b.py

where a.py is the app, and a.py has to import b.
In a.py you say I should use:
from my_project import b
Then I would need to do what to run the app:

cd ./<some-directory>
python3 <some-command>

[proud nebula]

No cd. You are in project/ and you type pytest. Hit enter.

No config file.

[tacit crown]

My project no longer runs with the changed imports.

[proud nebula]

You haven't said at all what you are doing. No one can help you if you don't answer questions about what you're doing.

[tacit crown]

What have you asked that I haven't answered?

[proud nebula]

"a.py" isn't "an app". It's a file. Your "app" is probably my_project.a at this point

[proud nebula]

"what did you do to get the error?"

also, you should get in the habit of showing full errors

[uneven beacon]

also worth reading this to better understand the import mechanisms https://docs.python.org/3/library/sys_path_init.html
to summarize the relevant parts:

• sys.path dictates where to search for modules when doing absolute imports
• by default, this includes python's built-in modules and the site-packages directory, containing your installed packages
• when running a script normally (i.e. python main.py), the directory containing that script gets searched before the others
• when running a module as if it was a script (i.e. python -m pytest), the current working directory of your terminal gets searched first

given the above facts, there are two (conventional) ways to go about it depending on your project layout:

• in an src-layout, the src/ protects mypackage from being found unless it is installed, so the workflow would be to install the package first, then run your commands like normal: py src/ mypackage/ __init__.py a.py from .b import foo b.py tests/ test_a.py from mypackage import a ```sh
  pip install -e .

open an interactive console with mypackage imported

python -ic "import mypackage"
pytest- in a flat-layout, it is possible for you to import the package from the project root without installation, although it requires your scripts to search the project root for your package:py
mypackage/
init.py
a.py
from .b import foo
b.py
tests/
test_a.py
from mypackage import a sh
python -ic "import mypackage"

run pytest with -m so the current working directory (which includes mypackage)

is available to your tests

python -m pytest```

[tacit crown]

For clarity, I've created a minimal test project.

Test\
    my_project\
        a.py
        b.py

The contents of a.py:

from my_project import b

def main():
    print(b.my_add(2, 3))


if __name__ == '__main__':
    main()

The contents of b.py:

def my_add(x, y):
    return a + b

I want to run a.py.
My cwd is .../Test/

How should I run a.py ?

[proud nebula]

python -m my_project.a

You don't really "run a.py". You run the module my_project.a.

[tacit crown]

$ python3 -m my_project.a
Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "<full-path>/Test/my_project/a.py", line 14, in <module>
    main()
  File "<full-path>/Test/my_project/a.py", line 10, in main
    print(b.my_add(2, 3))
  File "<full-path>/Test/my_project/b.py", line 8, in my_add
    return a + b
NameError: name 'a' is not defined

[uneven beacon]

your parameters are named x and y

[tacit crown]

😢 🤣 You're right.

OK, so moving on, I now get 5, so I'll try adding a test...

[proud nebula]

Don't confuse him further :P

[uneven beacon]

^ sidenote, if you had ran python my_project/a.py, python would run the script without any context about the package it was put inside, so an import like import my_project would be treated as importing my_project/my_project.py

[proud nebula]

no it wouldn't

[uneven beacon]

sure it would

[proud nebula]

ugh

python is a such a mess

[tacit crown]

I've added to my Test project:

Test\
    my_project\
        tests\
            test_1.py

And in test_1.py:

import unittest

from my_package import a, b

class TestIt(unittest.TestCase):
    def test_01(self):
        print('Hello World')

    def test_02(self):
        self.assertTrue(True)

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

Then I try to run:

$ python3 -m unittest discover tests

which returns the error:

E
======================================================================
ERROR: test_1 (unittest.loader._FailedTest)
----------------------------------------------------------------------
ImportError: Failed to import test module: test_1
Traceback (most recent call last):
  File "/usr/lib/python3.10/unittest/loader.py", line 436, in _find_test_path
    module = self._get_module_from_name(name)
  File "/usr/lib/python3.10/unittest/loader.py", line 377, in _get_module_from_name
    __import__(name)
  File "/<full-path>/Test/tests/test_1.py", line 3, in <module>
    from my_package import a, b
ModuleNotFoundError: No module named 'my_package'


----------------------------------------------------------------------
Ran 1 test in 0.000s

FAILED (errors=1)

[proud nebula]

that makes no sense

why in gods name does python add the directory of the file FIRST in sys.path?!

[proud nebula]

You probably want to put tests one level up

[uneven beacon]

ive seen that trip up plenty of people who named their discord bot discord.py then tried writing import discord

[proud nebula]

I once worked on a project where you could import both project.foo and foo. That was a biiiiig mess.

[uneven beacon]

btw you named your directory my_project

[tacit crown]

Sorry guys, I'm rushing to try and keep up. I'll fix that...

import unittest

from my_project import a, b

class TestIt(unittest.TestCase):
    def test_01(self):
        print('Hello World')

    def test_02(self):
        self.assertTrue(True)

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

which returns:

$ python3 -m unittest discover -t tests
Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/usr/lib/python3.10/unittest/__main__.py", line 18, in <module>
    main(module=None)
  File "/usr/lib/python3.10/unittest/main.py", line 100, in __init__
    self.parseArgs(argv)
  File "/usr/lib/python3.10/unittest/main.py", line 124, in parseArgs
    self._do_discovery(argv[2:])
  File "/usr/lib/python3.10/unittest/main.py", line 244, in _do_discovery
    self.createTests(from_discovery=True, Loader=Loader)
  File "/usr/lib/python3.10/unittest/main.py", line 154, in createTests
    self.test = loader.discover(self.start, self.pattern, self.top)
  File "/usr/lib/python3.10/unittest/loader.py", line 346, in discover
    raise ImportError('Start directory is not importable: %r' % start_dir)
ImportError: Start directory is not importable: '/<full-path>/Test'

and it's the "-t" that's wrong...

$ python3 -m unittest discover tests
Hello World
..
----------------------------------------------------------------------
Ran 2 tests in 0.000s

OK

Looking good...

[proud nebula]

Now, if you want to switch to a src style directory you have a baseline to work from here.

[uneven beacon]

i havent used unittest that much but according to the docs, it seems like you'd want the start directory to be tests but the top-level directory to be your project root (assuming a flat-layout where you havent installed the package)

[tacit crown]

import unittest

from my_project import a, b

class TestIt(unittest.TestCase):
    def test_01(self):
        self.assertEqual(b.my_add(5, 6), 11, "Should be 11")
    def test_02(self):
        self.assertEqual(b.my_add(6, 6), 11, "Should be 12")

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

$ python3 -m unittest discover tests
.F
======================================================================
FAIL: test_02 (test_1.TestIt)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/steve/Code/Python/Test/tests/test_1.py", line 9, in test_02
    self.assertEqual(b.my_add(6, 6), 11, "Should be 12")
AssertionError: 12 != 11 : Should be 12

----------------------------------------------------------------------
Ran 2 tests in 0.000s

FAILED (failures=1)

Hooray!
Thanks very much guys. I can take it from here.

[uneven beacon]

ah right -m is adding the project root for you

[tacit crown]

And interestingly, __init__.py was not required anywhere - so now to make a cup of tea and read up on Namespace packages.

Thanks again.

[rare atlas]

so I changed a program i made to use custom exceptions, both of which inherit from ValueError, and i ran pytest expecting to see a whole bunch of failing tests. But it turns out with pytest.raises(ValueError) also accepts ValueError subclasses. TIL

I don't want to in this case, but is there a way to prevent that?

[maiden pawn]

pytest captures the exception in an ExceptionInfo object. You can compare the exact type after the exception.

def test_get_holidays_raises_ioerror():
requests.get.side_effect = IOError
with pytest.raises(IOError) as excinfo:
get_holidays()
assert type(excinfo.value) is IOError
https://stackoverflow.com/questions/57160323/in-pytest-how-to-assert-if-an-exception-which-is-inherited-from-a-parent-excep

The answer is not wrong, but the additional assert has the same effect as an assert True. pytest.raises(IOError) will fail the test if another exception or no exception is raised, so the assert line won't be reached anyway. If you want to test for exceptions subclassing IOError, the assert will be reached, but will duplicate the type check already made in RaisesContext. –

[rare atlas]

that blows up differently that if you used a different Exception, but it does work. i actually think its complaint is more clear what the problem is, and it's certainly less verbose

[split field]

there are cases where you can get away without having a __init__.py, but you should have one, anyway. If you have a package that doesn't have a __init__.py, it's a namespace package - and if you don't know what a namespace package is, it's not what you meant to make. 🙂

[fiery arrow]

I read PEP420 way too many times and I think at this point my brain has developed immunity to understanding it 💀

[tacit crown]

It's not really a matter of "should" or "getting away with" __init__.py. Sorry, but it was because of dogmatic statements without adequate explanation, that I was confused in the first place. The presence or absence of __init__.py is largely irrelevant to the import problem that I was asking about.
Where __init__.py was relevant, was that support for Namespace packages in Unittest was broken for a long time before it was officially removed. During the years that it was broken, many "answers" were posted on-line for ways to work around the problem, frequently based on what worked in Python 2. The result: a confusing mess of bad and outdated advice and rules.
The real answer is as given in the Test Discovery section of the current Unittest documentation (https://github.com/python/cpython/blob/main/Doc/library/unittest.rst#test-discovery), which I would summarise as: everything must be importable from the top-level directory of the project.

[ember maple]

a caveat with pytest is that we have a pre-pep420 mechanism as of now, so __init__.py files are needed for sanity details

[digital pivot]

Is there any way to make pytest spawn a new process/reset everything for every test_module? Context - we are testing some applications which are in a monorepo, and all designed to run seperately from each other. We are using respx to mock endpoints, and finding that if we (for example) spawn a thread in module_1.py which was hitting a mocked api endpoint every few seconds, it could interfere with the tests in module_20.py.

There are things we can do to get around this, but the simplest would be to have pytest spawn a process/reset everything per test module. Other than just invoking pytest for each test_module, is there anyway around this?

[tall brook]

is it mock API or a fake API?

[proud nebula]

How are you mocking? Because normally the mocks clean up after themselves so you shouldn't get this problem.

[digital pivot]

using respx to setup the routes, with defined return values

[tall brook]

I haven't used respx but an idea could be to run each test in independent containers. Or perhaps dig deeper into respx docs and see how those scenarios are addressed

[ember maple]

there is pytest-forked but thats unmaintained and forks per test

nobody has spend the work on enabling such a use-case in xdist so far

[distant horizon]

Hi all! Does -s work anywhere that you might want to use log_cli=true? I.e. with -s is there any need for log_cli in the config?

[distant horizon]

I found that if you don't use with notation or decorators then you can make an unholy mess using = for mocking.

[ember maple]

its not quite clear what you are asking for there, live logging and stdio capture are different topics unless you do a footgun logging setup

[distant horizon]

@ember maple Is there any reason why log_cli=true would print to stdout but -s wouldnt? And if so, is there a cli option/arg to provide to enable log_cli=true?

[ember maple]

i don't quite follow - you can use -o log_cli=true to enable it via the cli, im not quite sure what you mean by the difference of it and -s do you perhaps do a own logging config as well

[distant horizon]

No the project I work on just treats everything at info level or above as a glorified printf and anything in stdout gets skurpeed up by the log collector.

We add some structure to it as well but we don't use any of the machinery that one might use to manage log levels for individual components.

[distant horizon]

We see a weird behaviour basically where in pycharm an improperly mocked celery task complains about not being able to access rabbit in the test if we use log_cli but not -s. And from the command line pytest cmd we can't get the same error to print even though we know it has the same problem as it's suffering from the same timeout behavior

[loud flame]

Hi all, I just discovered this discord. For background, most of my python code has been in Jupyter Notebooks (so please forgive my ignorance about unit testing tools in python), but I've recently needed to create a project that will be packaged and deployed with pip within my organization. The package uses an API for a cloud file storage/sharing platform, and I'm attempting do proper unit testing since it'll be used by others. I'm using pytest and unittest.mock for this and most things are testing well. I'm struggling with some of the mocking concepts related to how to set and assert mock objects attributes or method return values.

In my test code, I attempt to use @patch decorators to replace the external API calls. Specifically, there are two decorators in the testing scripts for each function. One for authentication to the API and another which returns a client object that should be returned by the authentication object's constructor. For example:

auth = APIAuth(someargs)
client = APIClient(auth)

So, I want to mock both the auth and the client objects for a test function. Within the test code, I can successfully do this: mock_auth.assert_called_with(someargs), but mock_client.assert_called_with(mock_auth) fails and shows that mock_auth's ID and the code under test's value for auth are different mock objects.

This is fairly important for my further testing, as the "client" object returned has many methods that are used in my program, and I need to be able to set various return values and side effects within that object to properly test things. So later in the program (and in other test scripts) this can happen:

folder = client.folder(folder_id=somearg)
newfile = folder.upload(somefilepath)
print('ID: ' + newfile.id)

And in my test, I want to set the return value of mock_client.folder().upload().id()'s return value to some bogus string and assert 'bogusstring' in result.output

Am I way off base in my efforts here? Could someone give me a hand or a nudge in the right direction? Thanks!

[proud nebula]

The first rule of monkey patching, is to not do monkey patching.

If you control the code, it's much better to pass in the thing into the function so you can control it. That removes the need for monkey patching to mock.

[loud flame]

I don't control the API's code. I'm trying to mock it so that I don't have to make real API calls to properly test.

...same as if I were trying to mock a database or other http request like so many unittest.mock examples around propose.

[earnest sequoia]

wrap the API in a proxy class. then you can mock your proxy class with another class that implements the same interface

[maiden pawn]

There is fearsomely too many times voiced word class 😁 despite being correct

[digital pivot]

Ended up doing something similar to this. Not containers - but running the tests in sub-processes and capturing the output seems to have done the job.

[sweet notch]

Is this crazy to have a fixture for my unit tests and fixture for integration tests that calls my unit test fixture?

@pytest.fixture(name="persisted_supplier")
def persisted_supplier(supplier_instance):
    db.drop_all()

    db.create_all()

    db.session.add(supplier_instance)

    db.session.commit()


# Unit fixtures
@pytest.fixture(name="supplier_instance")
def supplier_instance() -> Supplier:
    """
    Return a Supplier instance with sensible defaults for testing
    :return: Supplier
    """
    supplier = Supplier(
        id=1,
        name="dev_test_supplier_one",
    )
    return supplier```

idea is I call the supplier_instance for testing the model and then a persisted_supplier when I need to test the route.

[ember maple]

Not at all, but sometimes it's helpful to avoid classifying fixtures by strict test types and going for intent

What may be a footgun now is that the usage of the persisted fixture makes the fixture value of the supplier fixture persisted as well, but it does so at a distance

[sweet notch]

I need a supplier instance for model testing without persisting that data and then a persisted supplier for route testing.

That for the feedback here I need to do some thinking

I may just make both fixtures have no dependencies and instantiate a supplier in persisted_supplier function.

[ember maple]

Sometimes it makes sense to have small apis to create test data, and then using those in the fixtures

[sweet notch]

ahh I get you, have a utils file that will have functions that I can call from the fixture to create test data

I was debating looking in to factories also

[ember maple]

It's a good idea to find a proper name, typically utils modules are bags of things one didn't know where to put

[sweet notch]

Thanks for the direction dude, do you mind if I share my project with you later would love some feedback?

[slow vessel]

title: Mapped[str] = mapped_column(String(50), nullable=False, unique=True)
why does it let me insert empty rows even though nullable is false?

[ember maple]

Can't make promises for deep review, but i can do a glance

[sweet notch]

Thanks dude its a small project and I really appreciate your guidance!

will be sure to reflect what we have discussed in the project before I send

[spring veldt]

Hello from GoCodeo!
We have launched an AI-powered software that helps one write unit tests automatically for Python code.
Use our VS Code extension for free. Be a part of our alpha list now.
Website - www.gocodeo.com
Download VS Code extension - https://marketplace.visualstudio.com/items?itemName=GoCodeo.gocodeo

[ember maple]

i gave this a short test, the tests generated are extremely chatty with a high noise level and its easy to trigger hallucinations to get right looking wrong tests, im not impressed, im terrified what will happen if this gets overused by people not yet able to spot more subtle cases,

to add salt, i gave the "deceptive" code to chatgpt as well and it wrote correct unittests and explained why it ignored deceptive parts of the code

[spring veldt]

@ember maple we appreciate your feedback. We request you try our VS Code plug-in for full access and for the results you expect. The playground is open for basic codes and it is meant for demo purposes only.
VS Code plug-in is also available for free right now.

[maiden pawn]

@spring veldt i already tested a similar extension before and my final TLDR was next one

good - generates new ideas, like you have a partner in paired programming
bad - bad namings like novice programmer, not holding meaning why the test is needed
bad - too tightly coupled code tests to implementation details, making difficult continued development (your program becomes rigidly fixed to the current implementation, if u will try refactor it, you will start receive a lot of broken tests when they should not have been broken, potentially quite quickly leading to dissapointment of team of using tests in general)
bad - rips away thinking about code architecture of a program. Testing is important is exactly for this reason, to think about testable code architeture.
as you can see from this conclusion, this tool can get you short gain quickly (in covering current code with tests more), but the bad things... for majority of developers can rip away future of products they develop.

[spring veldt]

@maiden pawn we do not understand your feedback completely. From what we gather, it seems like you have a concern around the automated unit testing space. These platforms would be relevant to users who manually write a lot of unit tests. Our end goal is to assist them in reducing bugs and making their products stable.

[maiden pawn]

Copy pasting feedback in full way with context

[maiden pawn]

https://www.youtube.com/watch?v=NI-wYCQBOy0 original target of review

Hi there. I made sure to check your video and to analyze what you are offering as your tool.
Sorry for being late. A bit busy previous week i had. Hehe, was finishing my pet project too.

lets divide into what i like, what i dislike regarding your tool:

1. i like that your tool allows... to... generate a bunch of code, in different ways testing working solution.
   Surely great to generate ideas where you missed your code.
   I am not exactly sure only regarding... amount of spent time to carefully distinguish useful bits from not useful ones in this good spot.

2. i dislike that generated tests do not hold... mm meaning in test names, that holds why they are important?
   let me give you example. Imagine we asked intern-novice-junior programmer to comment what is this code doing?

for i in range(10):
  print(f"i={i}")

What we have high chance to receive?

this code does loop from 0 to 10, and prints i=values at each new string

for i in range(10):
  print(f"i={i}")

Absolutely exact description, and absolutely useless one. We don't know what this code doing, and comment just repeates what this code does in exactly repeating each encountered syntax, variable name and language action.
I see that so far tests generate have pretty much similar value

3. i potentially dislike also next thing: essentially the purpose of TDD is not about writing tests.
   It is about making you thinking how to write code architecture in the way, that your code would be testable.
   I am afraid that using this tool potentially propagates bad practice of not thinking about it. Thus eliminating pretty much all meaning of TDD.
4. • i dislike next thing also: writing tests should be as important as writing main code of a working solution.
     each test should be documenting some domain knowledge of our tool, and also we need to make them in the way, that they would be as least breaking code as possible. Through writing tests... in carefully chosen.... places where tests preferably test only external behavior of code places, and leave open room for easy refactoring code with the least amount of test fixes that would be needed after that.
     Tests written in the way they break as least as possible, allow us to write things with relience on written tests, and being confident we can easily fix them to next close commit, having them all passing green and feeling safe regarding our code.

• Your tool facilitates overtesting of code implementation details in a very... over thorough manner too much close to implementation details. Thus making code rigid-fixed by all those generated tests. It can bring to developer disappointment, that all tests are very quickly breaking when he tries to add new features or refactor code to new state.
  ====
  So, 2 potential likes for things good to have, and 3 things i potentially dislike regarding it.
  To say to for sure more about it, i would need to implement small project with using the tool, but i don't have that much time to invest for now.

good - generates new ideas, like you have a partner in paired programming
good - potentially increasing Dev speed
bad - bad namings like novice programmer, not holding meaning why the test is needed
bad - too tightly coupled code tests to implementation details, making difficult continued development
bad - rips away thinking about code architecture of a program
as you can see from this conclusion, this tool can get you short gain quickly, but the bad things... for majority of developers can rip away future of products they develop.

[maiden pawn]

====
my concern is by automating writing unit tests, you make fragile tightly coupled unit testing code tied to implementation too much like this CodiumAI tool
And kill thinkinking about making code architecture more testable

when developers thinks about how to test his code, he refactors it to make code architecture more testable

if we are just having auto generated unit tests, we have majority of meaning behind unit testing killed as well

[spring veldt]

@maiden pawn Thanks for the elaborated explanation. We are planning to provide certain customisation settings to allow developers feed in their inputs and insights. This will ensure that the testable architecture is in place. We also plan to launch settings that would allow refactoring based on the inputs of a developer.
Currently, our tool is focusing on reducing the manual effort of developers.

[spring veldt]

A simple code like the example you've provided wouldn't need an AI platform's assistance. They can be tested manually with ease.
Our product's use case is justified for complex codes that technology teams work on.

[ember maple]

How exactly should anyone trust something that makes glaring mistakes even in simple case not to make non obvious grave mistakes in hard cases,

[spring veldt]

@ember maple We don't think the platform has made any mistakes while creating unit tests for both simple and complex codes. This is as per our rigorous testing and user feedback from various other communities. In case you have found any mistakes, I would request you to share the screenshots of the results given by our machine.

[proud nebula]

No one owes you free QA :P

[spring veldt]

@proud nebula Haha. Why not?
We are following the footsteps of all great companies here.
Feedback <-> Continuous Improvement!

[ember maple]

Based on my initial tests, the feedback and the replies it's not worth my time to do a free quality qa for them,

[cerulean cradle]

what are the dependencies for unit testing?

new to python*

[brazen needle]

pytest is popular.

[cerulean cradle]

what is a test case id

[raven narwhal]

69

[spring veldt]

Thanks for trying our platform @ember maple

[drifting sorrel]

I would like to share a workflow I use on a daily basis when writing Python code: REPL Driven Development. I don't think that it is a common thing among Python developers (yet), but think it is very useful for developing testable code (and a good starting point for writing unit tests). If you haven't heard about it before: it is not about firing up a REPL and writing or importing code or scripts in there. Instead, you evaluate variables, code blocks, functions - or an entire module - without leaving your favorite code editor. The good part is the instant feedback. I use Emacs, and the elpy package has a good amount of features already in place for this kind of workflow - in combination with the powers of IPython. I have also tried this workflow out in PyCharm and VS Code with success. I have written about it, and you'll find more details here: https://davidvujic.blogspot.com/2022/08/joyful-python-with-repl.html.

[maiden pawn]

i can see some joy in it, but it should not be used as alternative to TDD
better write explorative unit tests, even if u will delete them later.

i can see it is as a good poor man choice if person hasn't learned how to do unit testing yet though

after all, Jypiter with its ability to write in its embedded repl exists too

Like.. there is no Point to specify Repl Driven Development, because during unit testing writing in IDE, we already use it (It is just called enabling interactive visual debug), and we already have this ability to execute code on a fly in Repl, except all code is already initialized for specific situation

[drifting sorrel]

My view of it is more like Test Driven Development Deluxe - with very fast feedback loops. A common thing is to actually write some "scratch style" code next to the actual code to test. These ones are easily converted into proper unit tests if you think it add value to keep.

[maiden pawn]

i think it is just called using interactive debug during test writing. It has same ability to execute any code too with instant feedback, except everything to prepare for this situation is documented as a code and u make very small necessary step forward to complete solution

[drifting sorrel]

It's not really debugging. You evaluate the code, without the context switch of writing something in a separate REPL (or even a unit test).

[maiden pawn]

you mean, you build each time this context from scratch? :/

[drifting sorrel]

Not following you there 🤔

[maiden pawn]

You evaluate the code, without the context switch of writing something in a separate REPL (or even a unit test).
so to execute the code, u need to initialize everything necessary for its execution from scratch

and u have bunch of arbitary code you don't know which (because not recorded as a code), that builds your repl environment

[drifting sorrel]

No, not at all. You have a session "alive" and you evaluate - and re-evaluate - code blocks as you type.

[maiden pawn]

ergh, essentially your suggestion is just using

if __name__=="__main__":
  my_scaffolding_code_i_write()

there is nothing new in this approach

it is a regular novice approach before they learn how to unit test code

[drifting sorrel]

hahaha, okay. I didn't say it is new, but probably not that heard of in the Python community.

[maiden pawn]

or if some very fast simple script is cooked up, too simple for proper written application code

[drifting sorrel]

I suggest you to read the blog post and check out the 5 minute video in there - that might clear things up for you a bit.

[maiden pawn]

technically it is not unit testing at all, and this discussion is not supposed to be there

we can say it is... a type of programming method. #software-architecture i guess

[drifting sorrel]

I believe it is about testing and according to the description in this channel I think it might be appropriate to share here.

[maiden pawn]

but it does not test code :/ tests are documentation to the code
your approach is equvalient to explorative tests which are written temporally, except your stuff is always deleted as a must

tests are supposed to be repeatable code written how to perform them

[ember maple]

Repl driven is good for experimentation, but it's very easy to miss necessary write the real tests cutoffs

And then it turns into self sabotage

[drifting sorrel]

Can you elaborate a bit on that? I don't see doing one or the other?

[ember maple]

A set of messy jupyter Notebooks ain't gonna turn themselves into a testsuite plus we'll factored code

[drifting sorrel]

Just to be clear: I would call myself a TDD:ist everyday of the week. Wrote my first Unit tests in 2005 correction: 2004 - and have not let go of it since.

I don't think REPL Driven Development is about Jupyter Notebooks - where does that come from?

[ember maple]

All other repl envs i tried had horrendous deceptions

In my experience a fast testsuite plus rerunning it gets more solid results

[drifting sorrel]

I think that you have misunderstood this thing - it's not about writing code in a REPL env. You would write the code in your editor, just as I described in the initial post, in the linked blog and demo in the 5 minute video you'll find in there.

[ember maple]

Python isn't exactly live coding save language, i had major pains with anything that pretends it can safely reveal partial snippets of code from the editor

[drifting sorrel]

I would encourage you then to try out REPL Driven Development - not much pain there.

And to be even extra clear, in case you think so: I am not suggesting anyone to replace unit tests or throwing away pytest at all.

[ember maple]

I just read your post, it's exactly the stuff that likes to break often

[drifting sorrel]

Have you even tried that out? I have developed Python for a bunch of years, and been practicing REPL Driven Development with the combination of writing unit tests for about 2-3 years.

I have yet not seen code break that often 😏

This kind of thing has helped me write code with better quality, and in combination with unit tests (especially TDD-style) it is a nice tool to have in your toolbox as a developer. I wanted to share that.

[ember maple]

I see some use in quick feedback for glue code, but any time one changes types or methods of stuff one has instances of, things get ugly quickly

[drifting sorrel]

Yeah, just as Unit Tests quickly become really ugly if you skip the refactoring part of your code. Guess most of the folks here have seen tests with endless mocking and setting up. No difference there.

[ember maple]

Yeah, without refactoring green eventually turns yellow 🙃

[drifting sorrel]

Totally agree 👆

[ember maple]

Also wrt mocks/mocking, massive amounts of it is toxic (so many people mock things they neither own nor understand)

https://hynek.me/articles/what-to-mock-in-5-mins/

[drifting sorrel]

I recognize that, and when reflecting on those kind of tests I think a huge part is production code that does too much - and the unit tests with those huge amounts of mocks become more of integration tests. That's where TDD and RDD can do great things. So, in a way TDD (and RDD) is more about software design than actual testing.

[ember maple]

Hmm, if we consider this as a spectrum between running the testsuite, running a debugger on failed tests and running code examples for exploration, there may be some better to agree on middle ground

[ember maple]

IMHO one of the issues that leads to such production code is that the starting point is something like a pseudo repl on top of auto restarting servers, i observed the drive towards growing them too much explorative, then writing bad tests on top that needs more mocks and integration tests than desirable

[drifting sorrel]

My experience is quite the opposite (if I understand you correctly) - starting off writing code, and then writing some tests. That workflow will often lead to painful sessions of patching and mocking code. The RDD way I suggest is very much like TDD - you write code, evaluate, try out, refactor, re-evaluate ... with that workflow - that shouldn't be too awkward for test-driven developers at least - will increase the chances of ending up with production code that is testable and that doesn't require those nasty amounts of mocking.

[ember maple]

I think we can agree on the fact that messily growing something on top of something like the flask or django devserver with no way too sanely run smaller examples is very far removed from the repl style you describe

[drifting sorrel]

Sounds like we're on the same page there!

[ember maple]

In essence you describe a lisp style repl, which is most powerful on top of a language with safe code reload and symbolic references

And the example size is basically something one would drop into a test function or a jupyter notebook cell

[drifting sorrel]

In such an environment (Lisp) you actually modify a running program itself - a very powerful thing to have as a developer. NRepl is a very powerful tool and I would very much like to have something like that in Python too.

[ember maple]

I think we have established that for python there is a number of footguns for repl driven development, we just seem to be blessed with different hit rates

[drifting sorrel]

Why write something like that? 👆

[ember maple]

It would be nice to have a python language server that could do symbolic replacement in the targeted interpreter

[brazen needle]

I have used "REPL-drived development" quite successfully sometimes. I find it's most useful to me when I'm thinking, "What am I even doing here?"

(I use the vim-slime package for this.)

I see it as complementary to unit testing (and other kinds of testing). I generally find myself using it on very small snippets.

It is often the case that these small snippets become functions which get tested properly. That's a step that you can't ignore (unless you're writing a throwaway script). At some point, it becomes more useful to write proper unit tests for these functions.

But sometimes it happens that I have only a vague idea of what the function is supposed to do. Maybe I know it's some kind of utility function for a larger piece of code, and there might be more than one of these, and I'm not sure quite what responsibilities each should have. Trying to implement things can make a better architecture clearer. This is the sort of situation where I find myself using a REPL for development.

It lets me develop enough to feel confident that things are structured reasonably and that they handle some inputs correctly. It's not a finished product, but it's progress.

Some people like to use Jupyter notebooks for this sort of thing. I generally dislike notebooks; I find it too easy to create bugs by executing cells out of order. (I feel like notebooks can be excellent for writing tutorials, though. I have done that successfully.)

[cerulean cradle]

if name == 'main':
unittest.main()
print ("Everything pass")
it says name is not defined..
any suggestions ?

[maiden pawn]

__name__=="__main__"

[proud nebula]

https://unofficial-django-discord.github.io/rules/markup.html to learn how to format code snippets in discord

[maiden pawn]

!code

[bitter wadi]

Formatting code on discord

Here's how to format Python code on Discord:

```py
print('Hello world!')
```

These are backticks, not quotes. Check this out if you can't find the backtick key.

For long code samples, you can use our pastebin.

[proud nebula]

ah

..weird. Why can't I put a 👍 on your !code comment?

Funny though, the entire Discord window shakes to say I am not allowed to do that.

[wicked vigil]

Hi guys. I create a python library. It can really help about unit test writing.It's good to have a look. There is a link https://github.com/fmelihh/chatgpt-test-generator

[proud nebula]

Seems backwards. If you can't even write tests, how could you know if the tests ChatGPT hallucinates are correct? Writing tests yourself and letting ChatGPT generate the code and then fixing it, seems much more logical.

[novel canopy]

I must agree with boxed

[drifting sorrel]

I haven't tried it out, but think the idea of an AI assisting by writing unit tests is pretty cool. I see use cases as something similar to a pair-programming (Human-AI) or a code-vs-test pingpong session for developing code. I haven't looked into that ChatGPT stuff yet, but applications of it like this would make me more interested than chatting.

[pseudo ermine]

I want to unit test a method of a class that calls a bunch of GUI methods in its constructor. If a create an object from that class the programm will crash (Its a crash in native C code so I cant catch it).

I dont really need the constructor to be called anyways, is there a way to just test the method without calling the constructor?

[drifting sorrel]

How about extracting the contents of the method into something that isn’t wiring up state? A function, or a separate class?

[ember maple]

There is no sound way, the common recommended approach is to ensure that one can make instances within setting the world on fire

As mentioned extraction may be the easiest approach for better suggestions details on the gui toolkit is needed

[upper patio]

Hi there. I'm working on an desktop automation project, using PyWinauto library. Can you imagine a way to create unit tests for classes/functions whose purpose is interacting with a GUI? I'm bad at testing, but afaik, the approach is to separate the logic that performs concrete actions like for example searching the window to find a specific element and click it. I'm unable to even think of was to do that. Does it even sound realistic?

[solemn solar]

May I ask what tools/extensions do u usually use for REPL dev? I've seen some good ones for nvim while I didn't find anything useful for VSCode.

[drifting sorrel]

I use Emacs but tried it out for VS Code and think it works quite well there. There’s a VS Code section in this post, and in the comments you’ll find ideas on how to setup key commands: https://davidvujic.blogspot.com/2022/08/joyful-python-with-repl.html

[solemn solar]

Thanks, this is super helpful. Actually I think I've read ur post before, which is how I got to know about RDD in the first place. 👍

[drifting sorrel]

Wow, that was fun to hear about. Thank you for sharing!

[proud nebula]

Have you looked at reloadium?

[solemn solar]

👍 Would be great with VSCode support

[proud nebula]

Pycharm >> vscode :)

[solemn solar]

Hh they're all great tools, I use neovim as well, just trying to make the best use of them.

[drifting sorrel]

Very cool, I didn't know about that tool but I have used a similar one (as it seems) called "Live coding in Python" that exist for PyCharm and Emacs.

[oblique shale]

!rule ad

[bitter wadi]

Rules

6. Do not post unapproved advertising.

[proud nebula]

<@&831776746206265384>

[minor furnace]

remove this

thanks @proud nebula

[crude marten]

I'd like to make a correction of this package : https://github.com/LAMDA-NJU/Deep-Forest
I tried to run test with pytest which returned me a first error :

Hint: make sure your test modules/packages have valid Python names.
Traceback:
/opt/homebrew/Cellar/python@3.11/3.11.2_1/Frameworks/Python.framework/Versions/3.11/lib/python3.11/importlib/__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
tests/test_tree_same.py:24: in <module>
    from sklearn.datasets import load_iris, load_wine, load_boston
/opt/homebrew/lib/python3.11/site-packages/sklearn/datasets/__init__.py:156: in __getattr__
    raise ImportError(msg)
E   ImportError: 
E   `load_boston` has been removed from scikit-learn since version 1.2.
E   
E   The Boston housing prices dataset has an ethical problem: as

I corrected it with fetch_california_housing not sure if it works for now.
But then I ran into another issue :

ImportError while importing test module '/Users/tools/Deep-Forest/tests/test_forest.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
/opt/homebrew/Cellar/python@3.11/3.11.2_1/Frameworks/Python.framework/Versions/3.11/lib/python3.11/importlib/__init__.py:126: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
test_forest.py:3: in <module>
    from deepforest import RandomForestClassifier
../deepforest/__init__.py:1: in <module>
    from .cascade import CascadeForestClassifier, CascadeForestRegressor
../deepforest/cascade.py:21: in <module>
    from . import _io, _utils
../deepforest/_utils.py:7: in <module>
    from . import _cutils as _LIB
E   ImportError: cannot import name '_cutils' from partially initialized module 'deepforest' (most likely due to a circular import) (/Users/dpratella/tools/Deep-Forest/deepforest/__init__.py)

I don't how to deal with this error.
I'm trying to work on the "test_forest.py"

Any help would be welcome !

And basically I just cloned it from github and ran pytest test_forest.py

[earnest sequoia]

You should open an issue on their GitHub

where the developers would actually be notified about it

[crude marten]

thought It would be faster for me to correct it by myself

[earnest sequoia]

oh, sorry I think I misunderstood your intention (when you said "correct", I thought you were "correcting" the devs xD)

you should check that your test script is run from the top level of the package

[crude marten]

I tried both :
pytest tests/test_forest.py
pytest test_forest.py

And still get the same message

[earnest sequoia]

hmm not sure then

[dense garnet]

This is how you do unit tests, right?
https://github.com/Adam-Vandervorst/PyBHV/blob/master/tests/laws.py

[wraith forge]

I'm having issues mocking reading a binary file if anyone is able to assist!

[proud nebula]

I vastly prefer pytest style teats over that.

[proud nebula]

You didn't ask a question

[wraith forge]

mb started to write one and got distracted

I'm writing a unit test for a function that opens/reads a binary log file but have not been able to get the mocker.mock_open function to work

    def test_read_log(self, mocker: MockFixture):
        """Check reading logs"""

        data = bytes(b"\x01\x02\x03\x04")
        data += bytes("0", "utf-8")
        with mocker.patch("builtins.open", mocker.mock_open(read_data=data)):
            with open("tempfile", "rb") as tempfile:
                log = blp.read_polling_log(tempfile, data_header_offset=0)

        assert log[0].body_string == "empty"
        assert log[0].record_body_len == 0

here's my code, it throws an AttributeError: __enter__ at the with mocker.patch line but i'm not sure what that means

[proud nebula]

That's not how mocking works. You mock the symbol where it is imported.

It's unintuitive and brittle. That's why mocking should be avoided. Well. One good reason.

[wraith forge]

i mean if there is a better option i would be happy to do it instead but this is the only way i know of to make this test

[proud nebula]

Show the code you want to test

[wraith forge]

do you mind if i dm it to you?

[dense garnet]

That was just a rough sketch, cleaned it up a lot, how about this?
https://github.com/Adam-Vandervorst/PyBHV/blob/master/tests/laws.py#L25

[bitter wadi]

tests/laws.py line 25

def lattice(join): return [associative(join), commutative(join), idempotent(join)]```

[proud nebula]

How was that reply relevant?

[proud nebula]

Show it to the channel so anyone can help

[wraith forge]

I got it figured out but thanks

Avoiding mocking was the right move 👍

[dense garnet]

Not sure what you mean. Maybe you can elaborate on why you prefer pytest or how you'd do property based testing with it across implementations.

[proud nebula]

It's just way cleaner. No classes that are literally useless.

As for PBT, the very first example in the docs for hypothesis uses pytest style tests: https://hypothesis.readthedocs.io/en/latest/quickstart.html

[proud nebula]

What does "across implementations" mean? Implementations of what?

[dense garnet]

Look at the cleaned up example; no classes 🙂

[dense garnet]

I looked at Hypothesis but it doesn't support law/property side abstraction. Nor does it seem to abstract well with parametrize or similar.

[dense garnet]

Multiple implementations of the same protocol you want to test.

[fallow meteor]

can someone guide me simplest framework/library to itarate through xls file row by row and use these rows in data driven test website? im pretty much stuck right now as i cant figure out how to find element to fill text -> take row 1 from xls file -> send keys to form -> click action -> take another row of data from file -> fill form and click again action button, untill end of file

just need some good tut or guide, becouse there is so much ways to do it, and i need it simple, as im not a programmer at all actually

[rapid patrol]

seems like it should be simple enough with selenium and pandas?

look into reading in excel files with pandas, then selenium for interacting with the site

[fallow meteor]

everywhere i see solution to just read all the table contents, and im not sure how to implement that input method on website use rows 1 by 1, i manage to use pandas but my input tries to use all table contents at once

actually im not sure if i should use iteration inside fill form and click button function, or fill form and click inside iteration function

[rapid patrol]

why not iterate over the rows in the dataframe?

for _,row in df.iterrows():
  # do things with row variable

considering it's sequential, the fact that the iterrows method being slow shouldn't be a problem

[fallow meteor]

but will it serve one rowe to execute, and wait with next one for next call? dataframe is littebit confusing for me as tutorials shows specific usecases without reasonale explaination

[rapid patrol]

provided you write an asynchronous program, then yes

I imagine that's the problem you're facing, you're trying to run the df synchronously, where the web browser interaction is inherently asynchronous, you must wait between "attempts"

[fallow meteor]

so there is no way that i get varialbe named "input" with passed row 1 to it, then used on website element, then iterate through next row2, send it to same variable and us it on website, then go to row3 send it to variable make actions on website and so on?

ofcourse with element.click() between variable change

[rapid patrol]

yeah of course, you might want some sleeps, or await a response code being sent back as your determining factor for when to move onto the next row

if efficiency were a concern for the iterrows thing, this post covers most methods, and efficiencies of them which is useful https://ryxcommar.com/2020/01/15/for-the-love-of-god-stop-using-iterrows/

[fallow meteor]

efficiency is not a thing, actually i have to use some sleep times to navigate website between searches as it is search box and i have to wait for results to be fully loaded on website

then provide new search quoery from list and repeat, but list is pretty long so embeding it in code is not possible, thats why im going with xls file

i started thinking that pandas is overengineered for that task and not straight forward for me to implement as im not exerience with programming coding beyond SQLqueries

maybe openpyxl would be a better option for this

[lucid lily]

Doing unit tests in django and this line causes to return this string for some reason, not really breaking bug but it annoys me for hours now

@classmethod
    def setUpTestData(self):
        user = User.objects.create(
            username="TestUser", password="VeryComplexFakePassword123[[]]")
        service = Service.objects.create(
            service_name="netflix", whole_service_name="Netflix", manager=user)
        group = PayingGroup.objects.create(
            group_name="My Netflix Group 1", monthly_cost="2", service=service, manager=user
        )
        subscriber = Subscriber.objects.create(
            name="Felix", iban="NL61RABO4110487447", debt="0", manager=user
        )
        subscriber.group.add(group)
        date = datetime.date(year=2023, month=3, day=29)
        #line under :/
        Payment.objects.create(
            subscriber=subscriber, date_of_payment=date, paid_amount="4", manager=user
        )

Found 4 test(s).
Creating test database for alias 'default'...
System check identified no issues (0 silenced).
Felix
....
----------------------------------------------------------------------
Ran 4 tests in 0.013s

OK
Destroying test database for alias 'default'...

**Returns Felix ** on line 4 of output

[lucid lily]

Here is Payment model object

class Payment(models.Model):
    subscriber = models.ForeignKey(
        Subscriber, on_delete=models.CASCADE, null=False)
    date_of_payment = models.DateField(
        auto_now_add=True)
    paid_amount = models.DecimalField(
        max_digits=6, decimal_places=2, default=0, null=False)
    manager = models.ForeignKey(
        to=User, on_delete=models.CASCADE, null=False)

    def __str__(self):
        return str(self.subscriber.name) + " " + str(self.paid_amount) + "€ " + str(self.date_of_payment)

[proud nebula]

"return this string"? You mean "prints"? And "this string" == "Felix"?

[proud nebula]

I mean.. you need to find the print() that is somewhere in your code base

[lucid lily]

yeah sorry for bad declaration

[proud nebula]

Can be annoying to find if you put a debugging print() somewhere and forgot about it for sure

[lucid lily]

I've tried to find it, but if I delete

Payment.objects.create(
            subscriber=subscriber, date_of_payment=date, paid_amount="4", manager=user
        )

problem goes away

[proud nebula]

That line triggers a LOT of code potentially

constructor of Payment, signals, just to say two obvious ones

[lucid lily]

now that you mention that I was tunnel visioned yeah I have signals connected to this model I will dig deeper thanks

[proud nebula]

Tunnel vision is the top reason to ask for help heh

[lucid lily]

@receiver(post_save, sender=Payment)
def UpdateTotalPaid(sender, instance, **kwargs):
    subscriber = Subscriber.objects.get(id=instance.subscriber.id)
    return print(subscriber.name)

thank you for saving me another hour of life

[proud nebula]

hah, that's funny, that'll just return None

[lucid lily]

xdd

[proud nebula]

hmm.. and not update total paid even

ouch

[lucid lily]

unfinished function

reworking app from admin only management to any logged in user can manage "dashboard" and some spiders and webs are left behind

[pastel idol]

hey
can anybody help me with pytest
how do i pass data from one test case to other
i have a function which traverses through a file directory and it generates the list
how do i pass that data to parmeterize marker

so i have 3 tests
test1 = it sends request and it gets a response
test2 = uses this response to test another parameter
test3 = uses the same response to test another parameter
how do i save the response
and send it to other testcases

[ember maple]

i

[ember maple]

that sounds like a single test you want to report as 3, i recommend taking a look at subtests instead of breaking test isolation as a starting point

[proud nebula]

Or maybe you make 1 test to check that the response is what you expect. Then the second test assumes the response you had expected. That way they are separate and can be run independently and parallel

[pastel idol]

So I have fixture which loops through all the files in a directory and creates a list

I have to pass this data to parameterize marker. How do I do it?
Can I use some other thing to do it?

[pastel idol]

Test case 2 needs data from test case 1 to execute
So test case 1,2,3 should happen sequentially

[proud nebula]

I reject that statement

[ember maple]

Currently fixtures happen after parameterization

[pastel idol]

Thank you,
I’ll look into subtests,

[pastel idol]

Yes that’s the problem

[pastel idol]

??

I’m new to pytest and testing 🥲

[proud nebula]

Think about it. If you get the results of a function, and assert it's some value. Then you KNOW it's that value. So the other test can assume it:

def foo():
    return 5

def bar(f):
    return f + 4

def test_foo():
    assert foo() == 5

def test_bar()
    result_of_foo = 5
    assert bar(result_of_foo) == 9

[pastel idol]

The problem here is that the data keeps changing dynamically

So I can’t assume

It’s script that generates some values
I’ll have to test it

The output value keeps changing

So I assume I have 2 scripts
And both of these scripts are generating values
I’ll have compare these 2 scripts that they are generating the same value

[ember maple]

That sounds like your testing approach misfits

[pastel idol]

Test case 1 is to test if the first script executed successfully
Test case 2 is to test if the second script generated same output has first one

[pastel idol]

So how can I test it then?

Thank you for the help! @ember maple @proud nebula

[proud nebula]

That... seems bad for a host of other reasons!

[ember maple]

More context is needed

Some key details are how do they get data, can it be synchronised and can it be turned into smaller parts that are easier to test

[pastel idol]

so there is tool which generates some data.
we are collecting this data and displaying it on a web page
i need to check if the tool generated the data
and then validate if the data on web page and the tool generated data is same or not

[pastel idol]

why do you think so?

[proud nebula]

Because it makes reasoning about the code very difficult and implies you might have a mutability mud pile.

"Functional core - imperative shell" is a good way to live.

[thorn leaf]

Hi, I'm using unittest for the first time on a project. I've created a function that implements a discrete probability distribution, and I want to use scipy's chi squared test to demonstrate that the outcomes obey the expected distribution. I want to create a list of test inputs, and run them through a test_chi_squared method - how do I structure my code so that this method is called once for each test input in my list? (is this the right channel for this sort of question?)

[maiden pawn]

so you already have installed pip libraries for your functionality?

https://docs.pytest.org/en/6.2.x/parametrize.html best just use pytest parametrization then ^_^

[thorn leaf]

you mean like scipy and stuff? I have that

is there a way to do it with unittest instead of pytest?

[maiden pawn]

is there any reason why u choose unittest over pytest?

https://pypi.org/project/parameterized/ in unittest, it will still involve installing libraries for that. Best just to go straight to pytest.

[thorn leaf]

hmm ok

makes sense

currently im trying to wrangle my scipy.chisquared function

[maiden pawn]

there should be a way generating parametrized tests in unittest

i remember being able to do that

but kind of no point to do that.

[thorn leaf]

if i need to install extra libraries probably not worth it

you know what im too tired for this i should just go to bed and tackle it tomorrow

thanks for the help, and goodnight

i might be asking similar questions tmrw

[maiden pawn]

https://chris-lamb.co.uk/posts/generating-dynamic-python-tests-using-metaclasses
ergh, there is possible using metaclasses for example, generating dynamically tests for unittest without libraries
I can bet it can be made simpler in some way

but you should not go this way :/

just use pytest, what stops u from that

[thorn leaf]

i will switch to pytest

[thorn leaf]

aight i got parameterised testing working with pytest

[maiden pawn]

🥳

[pastel idol]

Hey!!

I have a test case which collects data from 5 different machines using fabric
As this process is taking long time

I thought I'll use a seperate thread for each machine

I wrote a function which collects all the data
But function has few fixed values and one value which varies for different machine

I tried using pool
But it was throwing pickling error

How should I do it?

[maiden pawn]

why do u need collecting this data

[pastel idol]

To test the data

I need the data to validate for each machine

[maiden pawn]

not very descriptive why u need it

[pastel idol]

So every machine has a script that generates data.

I need to check if the data on each machine is correct or not

So in order to verify I need the data

[maiden pawn]

okay, why not dumping generated data into database
and accessing data over database

[pastel idol]

We are not using any database

[maiden pawn]

is this script forbidden to use databases for some reason? ^_^

for which reason

what stops you from using database to communicate data over networks

[pastel idol]

This is just a testing framework

So I'm using any database for now

The no machines is also less so I need to collect the data from machines itself

[maiden pawn]

in which form data is generated? is it a file?

[pastel idol]

It's a file

[maiden pawn]

tempted to offer ansible-playbook. it is made to connect multiple machines in parallel, and getting files / configuring them

and it is made in python

but it is kind of too heavy overkill for this perhaps

[pastel idol]

Yeah I guess

But it looks interesting

[maiden pawn]

besides the size itself it would do the job pretty fine

already ready solution

[pastel idol]

But I am using fabric now

[maiden pawn]

ansible is way more developed solution exactly made for this goal ^_^

fabric is apperently solution that u need to adjust

if fabric is a sync library, then u need writing probably multithreading to do the job with it (or may be subprocesses)

thinking logically, fabrics actions aren't CPU heavy, which means mulththreading should be enough

all you need in theory launching many threads in parallel, and awaiting to collect result from them all

[maiden pawn]

from threading import Thread

def fetch_rates(base):
  response = requests.get(
  f"https://api.vatcomply.com/rates?base={base}"
  )
  response.raise_for_status()
  rates = response.json()["rates"]
  # note: same currency exchanges to itself 1:1
  rates[base] = 1.
  rates_line = ", ".join(
  [f"{rates[symbol]:7.03} {symbol}" for symbol in SYMBOLS]
  )
  print(f"1 {base} = {rates_line}")


def main():
  threads = []
  for base in BASES:
    thread = Thread(target=fetch_rates, args=[base])
    thread.start()
    threads.append(thread)

  while threads:
    threads.pop().join()

launch threads with arguments fpr specific machines, await, and get results

[maiden pawn]

example is copy pasted form Chapter 6 of python expert programming 4th edition

(of course you can use thread pool too)

solution without thread pool is more simple, u can make it first ^_^ (check thread pool solution in the book)

[pastel idol]

@maiden pawn

Thank you!!

Multithreading is working

[pastel idol]

@maiden pawn in which chapter is the thread pool solution

[maiden pawn]

^_^ you replied to msg, which already mentions chapter number

[pastel idol]

Oh cool
Thank you!!!

[pastel idol]

if i have to return some from the thread how do i get it?

[maiden pawn]

a lazy solution would be using some thread safe data structure, and reading it is thread safe

that is the most advantage of threads, memory is shared ^_^

if data structure is not safe, lock can be used to ensure that (list is probably thread safe, but lock can be added anyway)

from threading import Thread
import threading 
lock = threading.RLock()
results = []

def fetch_rates(base: str):
  with lock:
    results.append(base)

def main():
  BASES = [1,2,3]

  threads = []
  for base in BASES:
    thread = Thread(target=fetch_rates, args=[base])
    thread.start()
    threads.append(thread)

  while threads:
    threads.pop().join()

  print("finished")
  print(results)

if __name__=="__main__":
  main()

[maiden pawn]

this solution is certainly stable now ^_^

[pastel idol]

if the function return some value

how do i get it?

[maiden pawn]

i have no idea, google it ^_^ or read the book i gave you, there probably mentioned solution

[maiden pawn]

https://www.youtube.com/watch?v=TH8yT3doasI

[pastel idol]

do you a have a copy of it ?

i found a copy online

but in chapter 6 there is some other topic

[maiden pawn]

The right one looks like this

[pastel idol]

is it by packt?

[maiden pawn]

secretly speaking we have channel dedicated to #async-and-concurrency, highly likely it is monitored by devs who used more often concurrency in their life ^_^

[pastel idol]

thank you!!

i'll post the question there and see

[proud nebula]

TIL that you can write cursed python with 2 space indents

that's horrible

[maiden pawn]

copy pasted from book 😛

[proud nebula]

ugh, you can of course do 1 space indented.. that's the most evil shit

[proud nebula]

That alone is cause for serious alarm

[maiden pawn]

during copy pasting all indents were removed. i was lazy to write 4