How to perform type checking with the typing Python module?

Question

I am reading the typing module code and also looking into mypy to understand how it does type checking. Unfortunately for me, mypy builds a very smart tree with typed expressions that I still don't understand, and it's all based on static analysis.

I would like to implement a type check system that is dynamic (no static analysis) in Python. Assuming the function that performs type checks is called check_type, I want to accomplish the following:

>>> import typing
>>>
>>> check_type(1, int)
True
>>> check_type(1, float)
False
>>> check_type([1], typing.List[int])
True
>>> check_type([1], typing.List[float])
False
>>> check_type([1], typing.List[typing.Any])
True
>>> check_type((1,), typing.Tuple[int])
True

I thought about recreating the object type from its value, such as:

>>> get_type([1])
typing.List<~T>[int]

But this doesn't work with issubclass:

>>> issubclass(typing.List[int], typing.List[typing.Any])
False

I don't see a simple way to check types in Python without assuming a lot of things about the internals of the typing stdlib module (e.g., accessing __args__ or __tuple_params__).

How can I properly implement a check_type function that work for the cases previously listed? I am using Python 2.7.

Answer 1

You can easily get the very limited functionality that works correctly for the simple examples provided in your question:

import mypy.api

def check_type(value, typ):
    program_text = 'from typing import *; v: {} = {}'.format(typ, repr(value))
    normal_report, error_report, exit_code = mypy.api.run(['-c', program_text])
    return exit_code == 0

int_ = 1
str_ = 'a'
list_str_ = ['a']
list_int_ = [1]
tuple_int_ = (1,)

assert check_type(int_, 'int')
assert not check_type(str_, 'int')
assert check_type(list_int_, 'List[int]')
assert not check_type(list_str_, 'List[int]')
assert check_type(list_int_, 'List[Any]')
assert check_type(tuple_int_, 'Tuple[int]')

You can even do some more advanced stuff (for example, refer to the types that correspond to classes you defined in your program) by extending this code a bit, so that mypy gets to parse your entire source code as opposed to just the current line.

Alternatively, you might want to look at enforce or typeguard.