GitHub - davidfritzsche/pytest-mypy-testing: pytest plugin to test mypy static type analysis (original) (raw)

PyPI GitHub Action Status License License: MIT

pytest-mypy-testing — Plugin to test mypy output with pytest

pytest-mypy-testing provides apytest plugin to test thatmypy produces a given output. As mypy can be told to display the type of an expressionthis allows us to check mypys type interference.

Installation

python -m pip install pytest-mypy-testing

The Python distribution package contains an entry pointso that the plugin is automatically discovered by pytest. To disable the plugin when it is installed , you can use the pytest command line option -p no:mypy-testing.

Writing Mypy Output Test Cases

A mypy test case is a top-level functions decorated with@pytest.mark.mypy_testing in a file named *.mypy-testing or in a pytest test module. pytest-mypy-testing follows the pytest logic in identifying test modules and respects thepython_filesconfig value.

Note that pytest-mypy-testing uses the Pythonast module to parse candidate files and does not import any file, i.e., the decorator must be exactly named @pytest.mark.mypy_testing.

In a pytest test module file you may combine both regular pytest test functions and mypy test functions. A single function can be both.

Example: A simple mypy test case could look like this:

@pytest.mark.mypy_testing def mypy_test_invalid_assignment() -> None: foo = "abc" foo = 123 # E: Incompatible types in assignment (expression has type "int", variable has type "str")

The plugin runs mypy for every file containing at least one mypy test case. The mypy output is then compared to special Python comments in the file:

mypy Error Code Matching

The algorithm matching messages parses mypy error code both in the output generated by mypy and in the Python comments.

If both the mypy output and the Python comment contain an error code and a full message, then the messages and the error codes must match. The following test case expects that mypy writes out anassignment error code and a specific error message:

@pytest.mark.mypy_testing def mypy_test_invalid_assignment() -> None: foo = "abc" foo = 123 # E: Incompatible types in assignment (expression has type "int", variable has type "str") [assignment]

If the Python comment does not contain an error code, then the error code written out by mypy (if any) is ignored. The following test case expects a specific error message from mypy, but ignores the error code produced by mypy:

@pytest.mark.mypy_testing def mypy_test_invalid_assignment() -> None: foo = "abc" foo = 123 # E: Incompatible types in assignment (expression has type "int", variable has type "str")

If the Python comment specifies only an error code, then the message written out by mypy is ignored, i.e., the following test case checks that mypy reports an assignment error:

@pytest.mark.mypy_testing def mypy_test_invalid_assignment() -> None: foo = "abc" foo = 123 # E: [assignment]

Skipping and Expected Failures

Mypy test case functions can be decorated with @pytest.mark.skip and@pytest.mark.xfail to mark them as to-be-skipped and as expected-to-fail, respectively. As with the@pytest.mark.mypy_testing mark, the names must match exactly as the decorators are extracted from the ast.

Development

Changelog

Unreleased

v0.1.3 (2024-03-05)

v0.1.2 (2024-02-26)

v0.1.1

v0.1.0

v0.0.12

v0.0.11

v0.0.10

v0.0.9

v0.0.8

v0.0.7

v0.0.6 - add pytest 5.4 support

v0.0.5 - CI improvements