Testing

Test your code!

Pylint is very well tested and has a high code coverage. New contributions are not accepted unless they include tests.

Before you start testing your code, you need to install your source-code package locally. To set up your environment for testing, open a terminal outside of your forked repository and run:

pip install -e <forked_repo_dir_name>

This ensures your testing environment is similar to Pylint's testing environment on GitHub.

Pylint uses two types of tests: unittests and functional tests.

  • The unittests can be found in the /pylint/test directory and they can be used for testing almost anything Pylint related.

  • The functional tests can be found in the /pylint/test/functional directory. They are mainly used to test whether Pylint emits the correct messages.

Before writing a new test it is often a good idea to ensure that your change isn't breaking a current test. You can run our tests using the tox package, as in:

python -m tox
python -m tox -epy36 # for Python 3.6 suite only
python -m tox -epylint # for running Pylint over Pylint's codebase
python -m tox -eformatting # for running formatting checks over Pylint's codebase

It's usually a good idea to run tox with --recreate. This flag tells tox to redownload all dependencies before running the tests. This can be important when a new version of astroid or any of the other dependencies has been published:

python -m tox --recreate # The entire tox environment will be recreated
python -m tox --recreate -e py310 # The python 3.10 tox environment will be recreated

To run only a specific test suite, use a pattern for the test filename (without the .py extension), as in:

python -m tox -e py310 -- -k test_functional
python -m tox -e py310 -- -k  \*func\*
python -m tox --recreate -e py310 -- -k test_functional # With recreation of the environment

Since we use pytest to run the tests, you can also use it on its own. We do recommend using the tox command though:

pytest pylint -k test_functional

Writing functional tests

These are residing under /pylint/test/functional and they are formed of multiple components. First, each Python file is considered to be a test case and it should be accompanied by a .txt file, having the same name, with the messages that are supposed to be emitted by the given test file.

In the Python file, each line for which Pylint is supposed to emit a message has to be annotated with a comment in the form # [message_symbol], as in:

a, b, c = 1 # [unbalanced-tuple-unpacking]

If multiple messages are expected on the same line, then this syntax can be used:

a, b, c = 1.test # [unbalanced-tuple-unpacking, no-member]

You can also use # +n: [ with n an integer if the above syntax would make the line too long or other reasons:

# +1: [empty-comment]
#

If you need special control over Pylint's configuration, you can also create a .rc file, which can have sections of Pylint's configuration. The .rc file can also contain a section [testoptions] to pass options for the functional test runner. The following options are currently supported:

  • "min_pyver": Minimal python version required to run the test

  • "max_pyver": Python version from which the test won't be run. If the last supported version is 3.9 this setting should be set to 3.10.

  • "min_pyver_end_position": Minimal python version required to check the end_line and end_column attributes of the message

  • "requires": Packages required to be installed locally to run the test

  • "except_implementations": List of python implementations on which the test should not run

  • "exclude_platforms": List of operating systems on which the test should not run

Functional test file locations

For existing checkers, new test cases should preferably be appended to the existing test file. For new checkers, a new file new_checker_message.py should be created (Note the use of underscores). This file should then be placed in the test/functional/n sub-directory.

Some additional notes:

  • If the checker is part of an extension the test should go in test/functional/ext/extension_name

  • If the test is a regression test it should go in test/r/regression or test/r/regression_02. The file name should start with regression_.

  • For some sub-directories, such as test/functional/u, there are additional sub-directories (test/functional/u/use). Please check if your test file should be placed in any of these directories. It should be placed there if the sub-directory name matches the word before the first underscore of your test file name.

The folder structure is enforced when running the test suite, so you might be directed to put the file in a different sub-directory.

Running and updating functional tests

During development, it's sometimes helpful to run all functional tests in your current environment in order to have faster feedback. Run from Pylint root directory with:

python tests/test_functional.py

You can use all the options you would use for pytest, for example -k "test_functional[len_checks]". Furthermore, if required the .txt file with expected messages can be regenerated based on the the current output by appending --update-functional-output to the command line:

python tests/test_functional.py --update-functional-output -k "test_functional[len_checks]"

Writing unittest tests

Most other tests reside in the '/pylint/test' directory. These unittests can be used to test almost all functionality within Pylint. A good step before writing any new unittests is to look at some tests that test a similar funcitionality. This can often help write new tests.

If your new test requires any additional files you can put those in the /pylint/test/regrtest_data directory. This is the directory we use to store any data needed for the unittests.

Writing functional tests for configurations

To test the different ways to configure Pylint there is also a small functional test framework for configuration files. These tests can be found in the '/pylint/test/config' directory.

To create a new test create a new file with an unused name in the directory of that type of configuration file. Subsequently add a filename.result.json file with 'filename' being the same name as your configuration file. This file should record what the configuration should be compared to the standard configuration.

For example, if the configuration should add a warning to the list of disabled messages and you changed the configuration for job to 10 instead of the default 1 the .json file should include:

"functional_append": {
    "disable": [["a-message-to-be-added"],]
}
"jobs": 10,

Similarly if a message should be removed you can add the following to the .json file:

"functional_remove": {
    "disable": [["a-message-to-be-removed"],]
}

If a configuration is incorrect and should lead to a crash or warning being emitted you can specify this by adding a .out file. This file should have the following name name_of_configuration_testfile.error_code.out. So, if your test is called bad_configuration.toml and should exit with exit code 2 the .out file should be named bad_configuration.2.out. The content of the .out file should have a similar pattern as a normal Pylint output. Note that the module name should be {abspath} and the file name {relpath}.

Primer tests

Pylint also uses what we refer to as primer tests. These are tests that are run automatically in our Continuous Integration and check whether any changes in Pylint lead to crashes or fatal errors on the stdlib and a selection of external repositories.

To run the primer tests you can add either --primer-stdlib or --primer-external to the pytest command. If you want to only run the primer you can add either of their marks, for example:

pytest -m primer_stdlib --primer-stdlib

The external primer has been split up in two marks to speed up our Continuous Integration. You can run either of the two batches or run them both:

pytest -m primer_external_batch_one --primer-external # Runs batch one
pytest -m primer_external_batch_two --primer-external # Runs batch two
pytest -m "primer_external_batch_one or primer_external_batch_two" --primer-external # Runs both batches

The list of repositories is created on the basis of three criteria: 1) projects need to use a diverse range of language features, 2) projects need to be well maintained and 3) projects should not have a codebase that is too repetitive. This guarantees a good balance between speed of our CI and finding potential bugs.

You can find the latest list of repositories and any relevant code for these tests in the tests/primer directory.