TopolearnPlugins turn pytest into a full testing platform, and a few of them are in almost every serious project.
Installing a plugin package is usually enough: plugins register through the pytest11 entry point and pytest loads them automatically. pytest --trace-config or the header of the run lists active plugins; -p no:name disables one.
The plugins nearly every project meets:
- pytest-xdist:
pytest -n autodistributes tests across one worker process per CPU. Tests must be independent and must not share files, ports or database names; useworker_idortmp_pathto isolate them. Session fixtures run once per worker. - pytest-cov:
pytest --cov=mypkg --cov-report=term-missingmeasures line (and with--cov-branch, branch) coverage and shows uncovered lines.--cov-fail-under=90enforces a threshold in CI. - pytest-asyncio: runs
async deftests. Mark them with@pytest.mark.asyncio, or setasyncio_mode = "auto"in configuration to treat every async test and fixture as asyncio. Async fixtures use@pytest_asyncio.fixturein strict mode. - pytest-mock (
mockerfixture), pytest-randomly (random order, exposes hidden dependencies between tests), pytest-timeout, pytest-django and anyio's plugin for trio/asyncio are other common picks.
Pin plugin versions alongside pytest, and use required_plugins in configuration so a missing plugin fails loudly instead of silently skipping behavior.
Example
# test_async.py (pip install pytest-asyncio pytest-xdist pytest-cov)
import asyncio
import pytest
async def fetch_price(symbol):
await asyncio.sleep(0.01) # pretend network call
return {"ACME": 12.5}.get(symbol)
@pytest.mark.asyncio
async def test_fetch_price():
assert await fetch_price("ACME") == 12.5
@pytest.mark.asyncio
async def test_concurrent_fetches():
prices = await asyncio.gather(fetch_price("ACME"), fetch_price("NOPE"))
assert prices == [12.5, None]
# $ pytest -n auto # parallel across CPUs
# $ pytest --cov=shop --cov-report=term-missing --cov-fail-under=90Remember
- Installed plugins load automatically via entry points
- pytest-xdist -n auto runs tests in parallel; isolate shared resources
- pytest-cov measures coverage and can enforce a minimum
- pytest-asyncio runs async def tests (marker or asyncio_mode=auto)
Watch out for
- Turning on -n auto with tests that share a fixed database name or temp file and getting random failures.