Information about testing

[mirpedrol]

Hi @tconbeer!

I am working on adding a TUI to our package nf-core/tools, currently implementing the snapshot tests.

I see your tests are asyncio events and then you are able to generate a snapshot with [app_snapshot()](https://github.com/tconbeer/harlequin/blob/974d8f7b922aa1b4b4b4d95f4309c67a7b060121/tests/functional_tests/test_app.py#L62). Could you point me to how to use this same approach?
When I try to add it to my tests I get an error fixture 'app_snapshot' not found

Thank you very much for your help!

[tconbeer]

app_snapshot is a pytest fixture that is defined by the pytest-textual-snapshot test dependency in Harlequin. I use my fork of Textualize's package. You need to install that package as a dependency before using the fixture; I don't recommend installing from my repo without forking it first, since I might change that at any time for Harlequin's purposes and break your stuff.

Basically, it uses Textual's built-in screenshotting capabilities and stores the SVG screenshot as a data snapshot with Syrupy. If the SVG changes between test runs, the test fails.

Textualize's package uses a fixture called snap_compare, which takes the path to an App file and a sequence of keystrokes. My fork adds the app_snapshot fixture, which you can call from inside a test script that uses the Pilot to drive the tests, which allows for multiple assertions per test setup.

These tests are easy to get started with, but frankly a pain to maintain. They are super brittle (changing a single pixel on the screen will fail the test -- sometimes the changes aren't even visible once the browser renders the SVG, since they're to hidden elements, etc.). They also execute pretty slowly. The tradeoffs are worth it to me, because I really care about quality, but they significantly slow down development.

[mirpedrol]

Thank you @tconbeer, this was very useful