SupercovCompare
← All comparisons

Python · Comparison

pytest-cov vs coverage.py

pytest-cov is a pytest plugin that runs coverage.py, so the numbers match. The real difference is pytest-xdist: plain coverage run reported 0% of lines.

pytest-cov or coverage.py: which should you run?

They do not compete: pytest-cov is a pytest plugin that runs coverage.py for you, so the numbers are the same. Use pytest-cov when you run tests in parallel with pytest-xdist, or want coverage as a pytest flag. Run coverage.py directly when you want one dependency fewer, or your runner is not pytest.

The two commands

pip install coverage
coverage run --branch --source=src -m pytest
coverage report -m
pip install pytest-cov
pytest --cov=src --cov-branch --cov-report=term-missing

Both read the same configuration: [tool.coverage] tables in pyproject.toml, or .coveragerc. Exclusions, fail_under and contexts apply either way. For the patterns that omit files and lines, see pytest coverage exclude.

Same numbers, one real difference

One module and three tests, on coverage.py 7.16.2, pytest-cov 7.1.0, pytest 9.1.1 and Python 3.13.1:

CommandLinesBranches
coverage run -m pytest6/62/2
pytest --cov6/62/2
coverage run -m pytest -n 20/6none
pytest -n 2 --cov6/62/2

With -n 2, pytest-xdist runs the tests in worker processes. Plain coverage run measured only the process that started them and warned “No data was collected”. pytest-cov starts coverage in every worker and combines the results.

coverage.py can do it alone with two settings and one more command. This also gave 6/6:

[run]
parallel = True
patch = subprocess
coverage run -m pytest -n 2
coverage combine
coverage report

What pytest-cov adds, and what it dropped

Its documentation lists four things over coverage run -m pytest: xdist support, combining and reporting in one step, --cov-context=test to record which test ran each line, and the same sys.path as a plain pytest run.

Since 7.0.0 it no longer measures subprocesses your tests start. Use coverage.py’s patch = subprocess for that. It needs coverage.py 7.10.6 or newer.

What neither shows

Both report 100% on this module. Two things were never tested:

  • if signed_in and not expired ran both ways, but no test has a signed-in session that has expired.
  • return user["name"] if user else fallback never returned fallback. coverage.py measures a branch as a line that can jump to more than one next line, and this line has one.

What Supercov adds

npx supercov -- pytest
npx supercov runs latest gaps

Supercov runs your own pytest command in your virtual environment, xdist workers included. On the same module it reported 6/6 lines, 5/6 branches and 1/3 MC/DC conditions, and named both gaps. It records coverage per test.

For other Python tools, see Python coverage: coverage.py, SlipCover, and Supercov.

Example evidence: fixture source, commands and recorded results (JSON). Measured October 12, 2026 with Supercov 4.3.0. These small examples are not a performance benchmark or a general conformance test.