Optional Pylint checkers in the extensions module

Parameter documentation checker

If you document the parameters of your functions, methods and constructors and their types systematically in your code this optional component might be useful for you. Sphinx style, Google style, and Numpy style are supported. (For some examples, see https://pypi.python.org/pypi/sphinxcontrib-napoleon .)

You can activate this checker by adding the line:

load-plugins=pylint.extensions.docparams

to the MASTER section of your .pylintrc.

This checker verifies that all function, method, and constructor parameters are mentioned in the

  • Sphinx param and type parts of the docstring:

    def function_foo(x, y, z):
        '''function foo ...
    
        :param x: bla x
        :type x: int
    
        :param y: bla y
        :type y: float
    
        :param int z: bla z
    
        :return: sum
        :rtype: float
        '''
        return x + y + z
    
  • or the Google style Args: part of the docstring:

    def function_foo(x, y, z):
        '''function foo ...
    
        Args:
            x (int): bla x
            y (float): bla y
    
            z (int): bla z
    
        Returns:
            float: sum
        '''
        return x + y + z
    
  • or the Numpy style Parameters part of the docstring:

    def function_foo(x, y, z):
        '''function foo ...
    
        Parameters
        ----------
        x: int
            bla x
        y: float
            bla y
    
        z: int
            bla z
    
        Returns
        -------
        float
            sum
        '''
        return x + y + z
    

You’ll be notified of missing parameter documentation but also of naming inconsistencies between the signature and the documentation which often arise when parameters are renamed automatically in the code, but not in the documentation.

By convention, constructor parameters are documented in the class docstring. (__init__ and __new__ methods are considered constructors.):

class ClassFoo(object):
    '''Sphinx style docstring foo

    :param float x: bla x

    :param y: bla y
    :type y: int
    '''
    def __init__(self, x, y):
        pass

class ClassFoo(object):
    '''Google style docstring foo

    Args:
        x (float): bla x
        y (int): bla y
    '''
    def __init__(self, x, y):
        pass

In some cases, having to document all parameters is a nuisance, for instance if many of your functions or methods just follow a common interface. To remove this burden, the checker accepts missing parameter documentation if one of the following phrases is found in the docstring:

  • For the other parameters, see
  • For the parameters, see

(with arbitrary whitespace between the words). Please add a link to the docstring defining the interface, e.g. a superclass method, after “see”:

def callback(x, y, z):
    '''Sphinx style docstring for callback ...

    :param x: bla x
    :type x: int

    For the other parameters, see
    :class:`MyFrameworkUsingAndDefiningCallback`
    '''
    return x + y + z

def callback(x, y, z):
    '''Google style docstring for callback ...

    Args:
        x (int): bla x

    For the other parameters, see
    :class:`MyFrameworkUsingAndDefiningCallback`
    '''
    return x + y + z

Naming inconsistencies in existing parameter and their type documentations are still detected.

By default, omitting the parameter documentation of a function altogether is tolerated without any warnings. If you want to switch off this behavior (forcing functions to document their parameters), set the option accept-no-param-doc to no in your .pylintrc.

By default, omitting the exception raising documentation of a function altogether is tolerated without any warnings. If you want to switch off this behavior (forcing functions that raise exceptions to document them), set the option accept-no-raise-doc to no in your .pylintrc.

By default, omitting the return documentation of a function altogether is tolerated without any warnings. If you want to switch off this behavior (forcing functions to document their returns), set the option accept-no-return-doc to no in your .pylintrc.

Prohibit builtin checker

This used to be the bad-builtin core checker, but it was moved to an extension instead. It can be used for finding prohibited used builtins, such as map or filter, for which other alternatives exists.

If you want to control for what builtins the checker should warn about, you can use the bad-functions option:

$ pylint a.py --load-plugins=pylint.extensions.bad_builtin --bad-functions=apply,reduce
...

Complexity checker

You can now use this plugin for finding complexity issues in your code base.

Activate it through pylint --load-plugins=pylint.extensions.mccabe. It introduces a new warning, too-complex, which is emitted when a code block has a complexity higher than a preestablished value, which can be controlled through the max-complexity option, such as in this example:

$ cat a.py
def f10():
    """McCabe rating: 11"""
    myint = 2
    if myint == 5:
        return myint
    elif myint == 6:
        return myint
    elif myint == 7:
        return myint
    elif myint == 8:
        return myint
    elif myint == 9:
        return myint
    elif myint == 10:
        if myint == 8:
            while True:
                return True
        elif myint == 8:
            with myint:
                return 8
    else:
        if myint == 2:
            return myint
        return myint
    return myint
$ pylint a.py --load-plugins=pylint.extensions.mccabe
R:1: 'f10' is too complex. The McCabe rating is 11 (too-complex)
$ pylint a.py --load-plugins=pylint.extensions.mccabe --max-complexity=50
$