llm-ide-rules: Instructions file for GitHub Copilot

Pytest Integration Tests

• Look to app/factories/ to generate any required database state
  • Here's an example of how to create + persist a factory DistributionFactory.save(domain=PYTHON_TEST_SERVER_HOST)
• Add the server factory to each test
• Use the faker factory to generate emails, etc.
• Don't add obvious assert descriptions
• Do not use the db_session fixture here. Instead, use with test_session(): if you need to setup complex database state
• if a UI timeout is occuring, it could be because it cannot find a element because the rendering has changed. Check the failure screenshot and see if you can correct the test assertion.
• The integration tests can take a very long time to run. Do not abort them if they are taking a long time.
• Use expect(page.get_by_text("Screening is fully booked")).to_be_visible() instead of expect(page.get_by_role("heading")).to_contain_text("Screening is fully booked"). It's less brittle.
• Do not use client fixtures in an integration test. Integration tests should only use the frontend of the website to interact with the application, not the API.
• Use with page.expect_response("https://example.com/resource") as response_info: to assert against network activity.
• Do not next_button.evaluate("el => el.click()") instead, just locator.click(). If this doesn't work, stop your work and let me know.
• Only use wait_for_loading(page) if a LONG_INTEGRATION_TEST_TIMEOUT on an expectation does not work: expect(page.get_by_text("Your Matched Doctors")).to_be_visible(timeout=LONG_INTEGRATION_TEST_TIMEOUT)
  • LONG_INTEGRATION_TEST_TIMEOUT should only be used as a last resort. If you have many of these in a test, let me know and I will debug it.
• Prefer fewer integration tests that cover more functionality. Unlike unit tests, where each test is designed to test a very particular piece of functionality, I want integration tests to cover entire workflows. It's preferred to add more steps to an integration test to test an entire workflow.
• Prefer simple locators. If a filter, or_, etc is required to capture a button in multiple states it indicates something is wrong in the code.
• Use react_router_url to generate the frontend url path and do not set base_url.
• End all Playwright tests with from pytest_playwright_artifacts import assert_no_console_errors and assert_no_console_errors(request) (capture is the plugin's playwright_console_logging fixture).
  • Test-Specific Ignores: Pass ignore=[...] to assert_no_console_errors per pytest-playwright-artifacts (regex strings, compiled patterns, or {"file": "...", "message": "..."} dicts); add a comment explaining why.
  • Global Ignores: Use playwright_console_ignore under [tool.pytest.ini_options] in pyproject.toml (see pytest-playwright-artifacts README).