How to use User Fixture in NiceGUI 3 -> unknown configuration value: 'main_file'

[BayerSe]

Hi,

I'm using the User Fixture to test my pages. With NiceGUI 2.24.2, the following setup works well.

# pyproject.toml
[project]
name = "nicegui-test"
version = "0.1.0"
requires-python = ">=3.13"
dependencies = [
    #"nicegui>3",
    "nicegui<3",
    "pytest>=8.4.2",
    "pytest-asyncio>=1.2.0",
]

[tool.pytest.ini_options]
asyncio_mode = "auto"

# main.py
from nicegui import ui

def create_pages():
    @ui.page('/')
    def main():
        ui.label('Hello World!')

# conftest.py
import pytest
from nicegui.testing import User
from main import create_pages

pytest_plugins = ["nicegui.testing.user_plugin"]

@pytest.fixture
def user(user: User) -> User:
    create_pages()
    return user

# test_app.py
from nicegui.testing import User

async def test_app(user: User) -> None:
    await user.open('/')
    await user.should_see('Hello World!')

With NiceGUI 3.0.2, I'm getting the following error:

ERROR test_app.py::test_app - ValueError: unknown configuration value: 'main_file'

I've seen that there is a main_faile mentioned here. If I add the following pytest.init and install selenium as well,

# pytest.ini
[pytest]
asyncio_mode = auto
main_file = main.py
addopts = -p nicegui.testing.plugin

I'm getting another error:

E           RuntimeError:
E
E           You must call ui.run() to start the server.
E           If ui.run() is behind a main guard
E              if __name__ == "__main__":
E           remove the guard or replace it with
E              if __name__ in {"__main__", "__mp_main__"}:
E           to allow for multiprocessing.

I'm confused. How am I supposed to use the User Fixture with NiceGUI > 3?

[rodja]

I'm sorry that it is not working for you out of the box @BayerSe. But I wonder if that is really your complete main.py? Where is your ui.run?

[BayerSe]

Thanks for reaching out @rodja! The way I have currently structured my project is that I have a separate module that configures logging and runs the ui:

from nicegui import app, ui
from main import create_pages

def run_dashboard() -> None:
    ... # configure logging
    app.on_startup(create_pages)
    ui.run()

The run_dashboard is launched via a script in my pyproject.toml

[project.scripts]
run_dashboard = 'run_dashboard:run_dashboard'

[rodja]

Ah. With 3.0 you do not need to do the separation hack via create_pages anymore. You can keep your create_dashboard command if you need it. Just make sure that main_file from pytest.ini finds the ui.run. Simplest changes:

Delete your conftest.py

Set main.py to:

from nicegui import ui

@ui.page('/')
def main():
    ui.label('Hello World!')

def run_dashboard():
    ui.run()

if __name__ in {'__main__', '__mp_main__'}:
    run_dashboard()

In your pyproject.toml:

[project.scripts]
run_dashboard = 'main:run_dashboard'

And test_app.py:

async def test_app(user: User) -> None:
    await user.open('/')
    await user.should_see('Hello World!')

The pytest.ini can stay as before:

[pytest]
asyncio_mode = auto
main_file = main.py
addopts = -p nicegui.testing.plugin

[BayerSe]

Thanks, @rodja. I was able to restore my tests.

• I now need to main two separate run_dashboard functions: one that configures logging + runs the ui with specific settings (ui.run(show=False, log_config=None, ...)) and one that simply runs the app with defaults, as you mentioned above. No big issue, just a little inconvenient.
• Is there a way to use the User Fixture without having to install selenium? Before, I could import the user plugin only but now I have to import the screen plugin too, which requires selenium.
• The new way of testing the app is somewhat limiting: we maintain two separate dashboards in the same repo and currently only test one of them. With the main_file in the pytest options, we can only test one of the dashboards, if I'm not mistaken?
• I think I have to continue using create_pages functions, as otherwise I can not instantiate separate dashboards.

[rodja]

I now need to main two separate run_dashboard functions: one that configures logging + runs the ui with specific settings (ui.run(show=False, log_config=None, ...)) and one that simply runs the app with defaults

I do not now your code so it might be the simplest solution. But I wonder why show=False etc. is not working in the tests. And if there are things you want to behave different, you can probe for beeing in a pytest execution with 'PYTEST_CURRENT_TEST' in os.environ.

[BayerSe]

The full entry point is along the lines of

    setup_opentelemetry()
    app.on_startup(create_pages)
    ui.run(
        host=host,
        port=port,
        show=False,
        reload=False,
        uvicorn_logging_level="info",
        log_config=None,
        root_path=root_path,
    )

a lot stuff going on here, not sure if / how this would interfer with testing. Besides that, I don't like having test-related code in production code.

[rodja]

Is there a way to use the User Fixture without having to install selenium? Before, I could import the user plugin only but now I have to import the screen plugin too, which requires selenium.

Oh wow. We should allow that. Can you create a feature request? Looks like it slipped through 😬

[rodja]

It's simple: just use addopts = -p nicegui.testing.user_plugin in your pytest.ini and not addopts = -p nicegui.testing.plugin.

[rodja]

I created #5262 to make it clear in the docs.

[rodja]

The new way of testing the app is somewhat limiting: we maintain two separate dashboards in the same repo and currently only test one of them. With the main_file in the pytest options, we can only test one of the dashboards, if I'm not mistaken?

True. Currently, it can only be changed globally in pytest.ini or pyproject.toml. However, allowing a command-line option or marker to configure get_path_to_main_file would make for a great pull request. Would you like to create a PR?

[BayerSe]

I'll have a look and report back.

[BayerSe]

#5269