Testing and benchmarking#
Testing#
Running the Python test suite#
Install the following dependencies into the default Python installation:
pip install pytest-pyodide pytest-httpserver
pytest-pyodide is a pytest plugin for testing Pyodide
and third-party applications that use Pyodide.
See: pytest-pyodide for more information.
Install geckodriver or chromedriver and check that they are in your
PATH.To run the test suite, run
pytestfrom the root directory of Pyodide:
pytest
There are 2 test locations that are collected by pytest,
src/tests/: general Pyodide tests and tests running the CPython test suitepackages/*/test_*: package specific tests.
You can run the tests from a specific file with:
pytest path/to/test/file.py
Some browsers sometimes produce informative errors than others
so if you are getting confusing errors it is worth rerunning the test on each
browser. You can use --runtime commandline option to specify the browser runtime.
pytest --runtime firefox
pytest --runtime chrome
pytest --runtime node
Custom test marks#
We support custom test marks:
@pytest.mark.skip_refcount_check and pytest.mark.skip_pyproxy_check disable
respectively the check for JavaScript references and the check for PyProxies.
If a test creates JavaScript references or PyProxies and does not clean them up,
by default the tests will fail. If a test is known to leak objects, it is
possible to disable these checks with these markers.
Running the JavaScript test suite#
To run tests on the JavaScript Pyodide package using Mocha, run the following commands,
cd src/js
npm test
To check TypeScript type definitions run,
npx tsd
Manual interactive testing#
To run tests manually:
Build Pyodide, perhaps in the docker image
From outside of the docker image,
cdinto thedistdirectory and runpython -m http.server.Once the webserver is running, simple interactive testing can be run by visiting the URL:
http://localhost:<PORT>/console.html. It’s recommended to usepyodide.runPythonin the browser console rather than using the repl.
Benchmarking#
To run common benchmarks to understand Pyodide’s performance, begin by installing the same prerequisites as for testing. Then run:
PYODIDE_PACKAGES="numpy,matplotlib" make benchmark
Linting#
We lint with pre-commit.
Python is linted with ruff, black and mypy.
JavaScript, markdown, yaml, and html are linted with prettier.
C is linted with clang-format.
To lint the code, run:
pre-commit run -a
You can have the linter automatically run whenever you commit by running
pip install pre-commit
pre-commit install
and this can later be disabled with
pre-commit uninstall
If you don’t lint your code, certain lint errors will be fixed automatically by
pre-commit.ci which will push fixes to your branch. If you want to push more
commits, you will either have to pull in the remote changes or force push.