CI testing a textual app

[nschloe]

For smoke testing, I'd like to start up textual from pytest, e.g.,

TiptopApp.run()

However, this gives

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <_pytest.capture.DontReadFromInput object at 0x7fd0e02a9520>

    def fileno(self) -> int:
>       raise UnsupportedOperation("redirected stdin is pseudofile, has no fileno()")
E       io.UnsupportedOperation: redirected stdin is pseudofile, has no fileno()

.tox/py3/lib/python3.9/site-packages/_pytest/capture.py:229: UnsupportedOperation

I'd also need a way for shutting down the app after N seconds or so.

[1j01]

Instead of App.run, use App.run_test, which will run the app in headless mode, and thus avoid accessing stdin/stdout.
You can use run_test's context manager (with syntax support) to get a Pilot object that can be used to interact with the app or wait before exiting. The app is exited automatically when the with scope closes.

from your_app.main import YourApp

async def test_app_can_run():
    app = YourApp()
    async with app.run_test() as pilot:
        await pilot.pause(3)

To use async tests, you'll need to install pytest-asyncio and enable it in pyproject.toml like so:

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

(or it can be enabled in pytest.ini or with a decorator @pytest.mark.asyncio)

---

I also tried using App.run with headless=True, but it failed to detect errors:

from textual.pilot import Pilot
from your_app.main import YourApp

def test_app_can_run_sync():
    # DON'T USE; this fails to catch errors; I don't know why.
    app = YourApp()
    async def auto_pilot(pilot: Pilot[object]):
        await pilot.pause(3)
        app.exit()  # not needed with run_test
    app.run(headless=True, size=(80, 24), auto_pilot=auto_pilot)