Type checking tool for Python 3

user_1264920 0
Viewed 584

I am trying to run a command that gives some aggregated information on type checking, static code analysis, etc. in some Python source code provided as a directory. If such thing exists, then I would like to add it to a Makefile invoked during some CI pipelines to validate a code base.

I've created this dummy source code file with a runtime issue

File foo_sample.py

def foo_func(my_num):
    return 1 + my_num


def bar_func():
    foo_func("foo")


if __name__ == "__main__":
    bar_func()

It throws this runtime error:

TypeError: unsupported operand type(s) for +: 'int' and 'str'.

Now I've tried to detect this type error with various tools, but they all fail to find these kind of problems. See the commands in the following list:

  • pyflakes foo_sample.py,
  • flake8 foo_sample.py,
  • pylint foo_sample.py,
  • mypy --check-untyped-defs foo_sample.py,
  • prospector --strictness veryhigh foo_sample.py

I know I could detect these issues with unittest, but the code base is quite large and untested to some degree. We frequently see runtime errors and address them with a unit test or an integration test. However, we would like to detect them during the execution of a CI pipeline.

Is there a way to run these kind of checks before running the code itself?

How can I avoid to discover these errors at runtime?

mypy static-code-analysis pylint typechecking python-3.x
2 Answers

I found pytype useful in this scenario.

When invoking pytype broken_code.py on this code:

def foo_func(my_num):
    return 1 + my_num


def bar_func():
    foo_func("foo")

def run_func(my_func):
    my_func()


if __name__ == "__main__":
    bar_func()
    run_func(bar_func())

I correctly find in the output:

  • line 3, in foo_func: unsupported operand type(s) for +: 'int' and 'str' [unsupported-operands]
  • line 10, in run_func: 'NoneType' object is not callable [not-callable]

The command exits with an error so a Makefile invoking it could block the execution of a pipeline in a CI environment.

Also, those error checks like [unsupported-operands], etc. can be enabled or disabled with comments in the code, e.g.,

  • first disable the checks # pytype: disable=attribute-error
  • then enable the checks again # pytype: enable=attribute-error

See how to use these comments in the documentation, Error classes.

user_12649200

Undeclared types are considered to be of type Any, and are not type checked by mypy. A stricter configuration is necessary to make sure mypy forces you to set types. Specifically, you need disallow_untyped_defs, which should lead you to this result:

$ cat test.py 
def foo_func(my_num: int) -> int:
    return 1 + my_num


def bar_func() -> None:
    foo_func("foo")


if __name__ == "__main__":
    bar_func()
$ mypy test.py
test.py:6: error: Argument 1 to "foo_func" has incompatible type "str"; expected "int"
Found 1 error in 1 file (checked 1 source file)

You might also want disallow_any_generics and warn_return_any. Example configuration.

user_965880
Related