mr. Proper

Static Python code analyzer, that tries to check if functions in code are pure or not and why.

DISCLAIMER: this library is very experimental and has a lot of edge cases. Functions that mr. Proper marks as pure can be not pure, but they are usually cleaner than other functions.

Installation

pip install mr_proper

What mr. Proper check

1. that function has no blacklisted calls (like print) and blacklisted attributes access (like smth.count);
2. that function not uses global objects (only local vars and function arguments);
3. that function has al least one return;
4. that function not mutates it's arguments;
5. that function has no local imports;
6. that function has no arguments of forbidden types (like ORM objects);
7. that function not uses self, class or super;
8. that function has calls of only pure functions.

This list is not enought to say that function is pure and some points are quite controversial, but it's a nice start.

Example

Console usage:

    # test.py
    def add_one(n: int) -> int:
        return n + 1

    def print_amount_of_users(users_qs: QuerySet) -> None:
        print(f'Current amount of users is {users_qs.count()}')

$ mr_propper test.py
add_one is pure!
print_amount_of_users is not pure because of:
    it uses forbidden argument types (QuerySet)
    it calls not pure functions (print)
    it has no return

Usage inside Python code sample:

>>> import ast
>>> from mr_propper.utils import is_function_pure
>>> funcdef = ast.parse('''
    def add_one(n: int) -> int:
        return n + 1
''').body[0]
>>> is_function_pure(funcdef)
True
>>> is_function_pure(funcdef, with_errors=True)
(True, [])

Parameters

CLI interface:

• filepath: path to .py file to check (directories are not supported for now);
• --recursive: require inner calls to be pure for function pureness.

Code prerequisites

1. Python 3.7+;
2. Functions are fully type-annotated;
3. No dynamic calls (like getattr(sender, 'send_' + message_type)(message)).

Contributing

We would love you to contribute to our project. It's simple:

1. Create an issue with bug you found or proposal you have. Wait for approve from maintainer.
2. Create a pull request. Make sure all checks are green.
3. Fix review comments if any.
4. Be awesome.

Here are useful tips:

• You can run all checks and tests with make check. Please do it before TravisCI does.
• We use BestDoctor python styleguide.
• We respect Django CoC. Make soft, not bullshit.