anoctopus / st-testing-demo

Streamlit App Testing Demo

This repo demonstrates the capabilities and usage of Streamlit's new AppTest class, released in Streamlit 1.28. App testing enables developers to:

• Run your app as a headless script
• Inspect and make assertions about the output content via the DOM in object-attribute style
• Programmatically modify the input values on various widgets, re-run the app, and inspect the output

AppTest works well with unittest.TestCase and pytest or similar tools. The testing framework ships with core Streamlit (and in fact is used heavily in Streamlit's internal unit testing) - the functionality shown here can be installed from the provided whl.

The example is built on the existing sophisticated_palette app originally built by @syasini and hosted on Streamlit Community Cloud: https://sophisticated-palette.streamlit.app/

A simple example

A simple example that demonstrates the anatomy of a unit test and the key features of the unit testing framework

import unittest
from streamlit.testing.v1 import AppTest

class TestSuite(unittest.TestCase):
    def test_sidebar(self):
        """Simple test of the sidebar controlling the palettes rendered"""

        # Load and run the script from a file path
        at = AppTest.from_file("app.py").run()

        # No exceptions were rendered in the app output
        assert not at.exception

        # Five color pickers are rendered on the first run
        assert len(at.color_picker) == 5

        # You can also query a widget by key to modify it or check the value
        assert at.number_input(key="sample_size").value == 500

        # Set the value of the first number input in the sidebar
        # (palette size) to 2, and re-run the app
        at.sidebar.number_input[0].set_value(2).run()

        # Two color pickers are rendered in the second run
        assert len(at.color_picker) == 2

Read the full documentation here.

See test_app.py for the tests and some further explanation.

Run it using GitHub Codespaces

You can get started very quickly by launching this repo in GitHub Codespaces. The app and Codespaces walkthrough should load automatically after a few seconds.

Run it locally

I recommend installing to a venv and then running the test suite with pytest.

git clone https://github.com/AnOctopus/st-testing-demo.git
cd st-testing-demo
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pytest

Using it with CI

The testing framework just uses pytest, so any CI tools that work with python should just work. You can see a Github Action workflow in this repo that runs the tests using GitHub's python starter workflow with zero modifications, and it works great.

You can view the Github Actions configuration for this repository, and recent CI runs including the test execution.

🎈 Let us know what you think! 🎈

• 👉 Add your feedback in the forum