Testing in Python

pytest is the default for new projects: plain assert statements, fixtures for setup, parametrized tests and a large plugin ecosystem. It also runs tests written for the standard library's unittest. Hypothesis adds property-based testing: you describe the inputs, it generates many cases and shrinks any failure to a minimal example. Robot Framework uses keyword-driven tests that non-programmers can read, which suits acceptance testing. tox and nox run your suite across Python versions and environments; nox is configured in Python rather than INI.

For isolation, unittest.mock patches objects in the standard library, responses mocks the Requests library, RESPX mocks HTTPX, VCR.py records real HTTP interactions and replays them, and time-machine controls the current time. factory_boy and Polyfactory build test objects, with Polyfactory generating them from type hints on dataclasses and Pydantic models. Faker and Mimesis generate realistic fake data, and Coverage.py measures what your tests exercise.

For browsers, Playwright drives Chromium, Firefox and WebKit with automatic waiting, Selenium is the long-standing WebDriver standard, and SeleniumBase wraps Selenium with extra helpers. Locust writes load tests as Python code, and Schemathesis generates API tests from OpenAPI and GraphQL schemas. Playwright is the usual pick for new end-to-end suites because its automatic waiting removes most flaky sleeps, and it includes a pytest plugin. Selenium remains the choice when you need a WebDriver grid or browsers Playwright does not drive.