Skip to content


Repository files navigation

mypy logo

pytest plugin for testing mypy types, stubs, and plugins

Tests Status Checked with mypy Gitter PyPI Conda Version


This package is available on PyPI

pip install pytest-mypy-plugins

and conda-forge

conda install -c conda-forge pytest-mypy-plugins



Plugin, after installation, is automatically picked up by pytest therefore it is sufficient to just execute:


Asserting types

There are two ways to assert types. The custom one and regular typing.assert_type.

Our custom type assertion uses reveal_type helper and custom output matchers:

- case: using_reveal_type
  main: |
    instance = 1
    reveal_type(instance)  # N: Revealed type is ''

This method also allows to use # E: for matching exact error messages and codes.

But, you can also use regular assert_type, examples can be found here.


The PYTHONPATH and MYPYPATH environment variables, if set, are passed to mypy on invocation. This may be helpful if you are testing a local plugin and need to provide an import path to it.

Be aware that when mypy is run in a subprocess (the default) the test cases are run in temporary working directories where relative paths such as PYTHONPATH=./my_plugin do not reference the directory which you are running pytest from. If you encounter this, consider invoking pytest with --mypy-same-process or make your paths absolute, e.g. PYTHONPATH=$(pwd)/my_plugin pytest.

You can also specify PYTHONPATH, MYPYPATH, or any other environment variable in env: section of yml spec:

- case: mypy_path_from_env
  main: |
    from pair import Pair

    instance: Pair
    reveal_type(instance)  # N: Revealed type is 'pair.Pair'
    - MYPYPATH=../fixtures

What is a test case?

In general each test case is just an element in an array written in a properly formatted YAML file. On top of that, each case must comply to following types:

Property Type Description
case str Name of the test case, complies to [a-zA-Z0-9] pattern
main str Portion of the code as if written in .py file
files Optional[List[File]]=[]* List of extra files to simulate imports if needed
disable_cache Optional[bool]=False Set to true disables mypy caching
mypy_config Optional[str] Inline mypy configuration, passed directly to mypy as --config-file option, possibly joined with --mypy-pyproject-toml-file or --mypy-ini-file contents if they are passed. By default is treated as ini, treated as toml only if --mypy-pyproject-toml-file is passed
env Optional[Dict[str, str]]={} Environmental variables to be provided inside of test run
parametrized Optional[List[Parameter]]=[]* List of parameters, similar to @pytest.mark.parametrize
skip str Expression evaluated with following globals set: sys, os, pytest and platform
expect_fail bool Mark test case as an expected failure, like @pytest.mark.xfail
regex str Allow regular expressions in comments to be matched against actual output. Defaults to "no", i.e. matches full text.

(*) Appendix to pseudo types used above:

class File:
    path: str
    content: Optional[str] = None
Parameter = Mapping[str, Any]

Implementation notes:

  • main must be non-empty string that evaluates to valid Python code,
  • content of each of extra files must evaluate to valid Python code,
  • parametrized entries must all be the objects of the same type. It simply means that each entry must have exact same set of keys,
  • skip - an expression set in skip is passed directly into eval. It is advised to take a peek and learn about how eval works.

Repository also offers a JSONSchema, with which it validates the input. It can also offer your editor auto-completions, descriptions, and validation.

All you have to do, add the following line at the top of your YAML file:

# yaml-language-server: $schema=


1. Inline type expectations

# typesafety/test_request.yml
- case: request_object_has_user_of_type_auth_user_model
  main: |
    from django.http.request import HttpRequest
    reveal_type(HttpRequest().user)  # N: Revealed type is 'myapp.models.MyUser'
    # check that other fields work ok
    reveal_type(HttpRequest().method)  # N: Revealed type is 'Union[builtins.str, None]'
    - path: myapp/
    - path: myapp/
      content: |
        from django.db import models
        class MyUser(models.Model):

2. @parametrized

- case: with_params
    - val: 1
    - val: 1.0
      rt: builtins.float
  main: |
    reveal_type({{ val }})  # N: Revealed type is '{{ rt }}'

Properties that you can parametrize:

  • main
  • mypy_config
  • out

3. Longer type expectations

- case: with_out
  main: |
  out: |
    main:1: note: Revealed type is 'builtins.str'

4. Regular expressions in expectations

- case: expected_message_regex_with_out
  regex: yes
  main: |
    a = 'abc'
  out: |
    main:2: note: .*str.*

5. Regular expressions specific lines of output.

- case: expected_single_message_regex
  main: |
    a = 'hello'
    reveal_type(a)  # NR: .*str.*


                        Base directory for tests to use
                        Which `pyproject.toml` file to use
                        as a default config for tests.
                        Incompatible with `--mypy-ini-file`
                        Which `.ini` file to use as a default config for tests.
                        Incompatible with `--mypy-pyproject-toml-file`
                        Run in the same process. Useful for debugging,
                        will create problems with import cache
                        Fully qualified path to the extension hook function,
                        in case you need custom yaml keys. Has to be top-level
                        mypy will ignore errors from site-packages
                        Use closed schema to validate YAML test cases,
                        which won't allow any extra keys
                        (does not work well with `--mypy-extension-hook`)

Further reading