Best practices

These checks ensures that you follow the best practices.

The source for these best practices is hidden inside countless hours we have spent debugging software or reviewing it.

How do we find an inspiration for new rules? We find some ugly code during code reviews and audits. Then we forbid to use it forever. So, this error will never return to our codebase.

Summary

WrongMagicCommentViolation	Restricts to use several control (or magic) comments.
WrongDocCommentViolation	Forbids to use empty doc comments (#:).
WrongModuleMetadataViolation	Forbids to have some module level variables.
EmptyModuleViolation	Forbids to have empty modules.
InitModuleHasLogicViolation	Forbids to have logic inside __init__ module.
WrongKeywordViolation	Forbids to use some keywords from python.
WrongFunctionCallViolation	Forbids to call some built-in functions.
FutureImportViolation	Forbids to use __future__ imports.
RaiseNotImplementedViolation	Forbids to use NotImplemented error.
BaseExceptionViolation	Forbids to use BaseException exception.
BooleanPositionalArgumentViolation	Forbids to pass booleans as non-keyword parameters.
NestedFunctionViolation	Forbids to have nested functions.
NestedClassViolation	Forbids to use nested classes.
MagicNumberViolation	Forbids to use magic numbers in your code.
StaticMethodViolation	Forbids to use @staticmethod decorator.
BadMagicMethodViolation	Forbids to use some magic methods.
NestedImportViolation	Forbids to have nested imports in functions.
RedundantLoopElseViolation	Forbids to use else without break in a loop.
RedundantFinallyViolation	Forbids to use finally in try block without except block.
ReassigningVariableToItselfViolation	Forbids to assign variable to itself.
YieldInsideInitViolation	Forbids to use yield inside of __init__ method.
ProtectedModuleViolation	Forbids to import protected modules.
ProtectedAttributeViolation	Forbids to use protected attributes and methods.
LambdaInsideLoopViolation	Forbids to use lambda inside loops.
UnreachableCodeViolation	Forbids to have unreachable code.
StatementHasNoEffectViolation	Forbids to have statements that do nothing.
MultipleAssignmentsViolation	Forbids to have statements that do nothing.
IncorrectUnpackingViolation	Forbids to have statements that do nothing.
DuplicateExceptionViolation	Forbids to have the same exception class in multiple except blocks.

Comments

class WrongMagicCommentViolation(node=None, text=None)

Bases: wemake_python_styleguide.violations.base.SimpleViolation

Restricts to use several control (or magic) comments.

We do not allow to use:

1. # noqa comment without specified violations
2. # type: some_type comments to specify a type for typed_ast

Reasoning:

We cover several different use-cases in a single rule. # noqa comment is restricted because it can hide other violations. # type: some_type comment is restricted because we can already use type annotations instead.

Solution:

Use # noqa comments with specified error types. Use type annotations to specify types.

We still allow to use # type: ignore comment. Since sometimes it is totally required.

Example:

# Correct:
type = MyClass.get_type()  # noqa: A001
coordinate: int = 10
some.int_field = 'text'  # type: ignore

# Wrong:
type = MyClass.get_type()  # noqa
coordinate = 10  # type: int

New in version 0.1.0.

code = 400

error_template = 'Found wrong magic comment: {0}'

class WrongDocCommentViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.TokenizeViolation

Forbids to use empty doc comments (#:).

Reasoning:

Doc comments are used to provide a documentation. But supplying empty doc comments breaks this use-case. It is unclear why they can be used with no contents.

Solution:

Add some documentation to this comment. Or remove it.

Empty doc comments are not caught by the default pycodestyle checks.

Example:

# Correct:
#: List of allowed names:
NAMES_WHITELIST = ['feature', 'bug', 'research']

# Wrong:
#:
NAMES_WHITELIST = ['feature', 'bug', 'research']

New in version 0.1.0.

code = 401

error_template = 'Found wrong doc comment'

Modules

class WrongModuleMetadataViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have some module level variables.

Reasoning:

We discourage using module variables like __author__, because code should not contain any metadata.

Solution:

Place all the metadata in setup.py, setup.cfg, or pyproject.toml. Use proper docstrings and packaging classifiers. Use pkg_resources if you need to import this data into your app.

See MODULE_METADATA_VARIABLES_BLACKLIST for full list of bad names.

Example:

# Wrong:
__author__ = 'Nikita Sobolev'
__version__ = 0.1.2

New in version 0.1.0.

error_template = 'Found wrong metadata variable: {0}'

code = 410

class EmptyModuleViolation(node=None, text=None)

Bases: wemake_python_styleguide.violations.base.SimpleViolation

Forbids to have empty modules.

Reasoning:

Why is it even there? Do not polute your project with empty files.

Solution:

If you have an empty module there are two ways to handle that:

1. delete it
2. drop some documentation in it, so you will explain why it is there

New in version 0.1.0.

error_template = 'Found empty module'

code = 411

class InitModuleHasLogicViolation(node=None, text=None)

Bases: wemake_python_styleguide.violations.base.SimpleViolation

Forbids to have logic inside __init__ module.

Reasoning:

If you have logic inside the __init__ module it means several things:

1. you are keeping some outdated stuff there, you need to refactor
2. you are placing this logic into the wrong file, just create another one
3. you are doing some dark magic, and you should not do that

Solution:

Put your code in other modules.

However, we allow to have some contents inside the __init__ module:

1. comments, since they are dropped before AST comes in play
2. docs string, because sometimes it is required to state something

It is also fine when you have different users that use your code. And you do not want to break everything for them. In this case this rule can be configured.

Configuration:

This rule is configurable with --i-control-code. Default: True

New in version 0.1.0.

error_template = 'Found `__init__.py` module with logic'

code = 412

Builtins

class WrongKeywordViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use some keywords from python.

Reasoning:

We believe that some keywords are anti-patterns. They promote bad-practices like global and pass, or just not user-friendly like del.

Solution:

Solutions differ from keyword to keyword. pass should be replaced with docstring or contextlib.suppress. del should be replaced with specialized methods like .pop(). global and nonlocal usages should be refactored.

Example:

# Wrong:
pass
del
nonlocal
global

New in version 0.1.0.

error_template = 'Found wrong keyword: {0}'

code = 420

class WrongFunctionCallViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to call some built-in functions.

Reasoning:

Some functions are only suitable for very specific use cases, we forbid to use them in a free manner.

See FUNCTIONS_BLACKLIST for the full list of blacklisted functions.

See also

https://www.youtube.com/watch?v=YjHsOrOOSuI

New in version 0.1.0.

error_template = 'Found wrong function call: {0}'

code = 421

class FutureImportViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use __future__ imports.

Reasoning:

Almost all __future__ imports are legacy python2 compatibility tools that are no longer required.

Solution:

Remove them. Drop python2 support.

Except, there are some new ones for python4 support. See FUTURE_IMPORTS_WHITELIST for the full list of allowed future imports.

Example:

# Correct:
from __future__ import annotations

# Wrong:
from __future__ import print_function

New in version 0.1.0.

error_template = 'Found future import: {0}'

code = 422

class RaiseNotImplementedViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use NotImplemented error.

Reasoning:

These two violations look so similar. But, these violations have different use cases. Use cases of NotImplemented is too limited to be generally available.

Solution:

Use NotImplementedError.

Example:

# Correct:
raise NotImplementedError('To be done')

# Wrong:
raise NotImplemented

New in version 0.1.0.

See also

https://stackoverflow.com/a/44575926/4842742

error_template = 'Found raise NotImplemented'

code = 423

class BaseExceptionViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use BaseException exception.

Reasoning:

We can silence system exit and keyboard interrupt with this exception handler. It is almost the same as raw except: block.

Solution:

Handle Exception, KeyboardInterrupt, GeneratorExit, and SystemExit separately. Do not use the plain except: keyword.

Example:

# Correct:
except Exception as ex: ...

# Wrong:
except BaseException as ex: ...

New in version 0.3.0.

See also

https://docs.python.org/3/library/exceptions.html#exception-hierarchy https://help.semmle.com/wiki/pages/viewpage.action?pageId=1608527

error_template = 'Found except `BaseException`'

code = 424

class BooleanPositionalArgumentViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to pass booleans as non-keyword parameters.

Reasoning:

Passing boolean as regular positional parameters is very non-descriptive. It is almost impossible to tell, what does this parameter means. And you almost always have to look up the implementation to tell what is going on.

Solution:

Pass booleans as keywords only. This will help you to save extra context on what’s going on.

Example:

# Correct:
UsersRepository.update(cache=True)

# Wrong:
UsersRepository.update(True)

New in version 0.6.0.

error_template = 'Found boolean non-keyword argument'

code = 425

Design

class NestedFunctionViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have nested functions.

Reasoning:

Nesting functions is a bad practice. It is hard to test them, it is hard then to separate them. People tend to overuse closures, so it’s hard to manage the dataflow.

Solution:

Just write flat functions, there’s no need to nest them. Pass parameters as normal arguments, do not use closures. Until you need them for decorators or factories.

We also disallow to nest lambda and async functions.

See NESTED_FUNCTIONS_WHITELIST for the whole list of whitelisted names.

Example:

# Correct:
def do_some(): ...
def other(): ...

# Wrong:
def do_some():
    def inner():
        ...

New in version 0.1.0.

error_template = 'Found nested function: {0}'

code = 430

class NestedClassViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use nested classes.

Reasoning:

Nested classes are really hard to manage. You can not even create an instance of this class in many cases. Testing them is also really hard.

Solution:

Just write flat classes, there’s no need nest them. If you are nesting classes inside a function for parametrization, then you will probably need to use different design (or metaclasses).

See NESTED_CLASSES_WHITELIST for the full list of whitelisted names.

Example:

# Correct:
class Some(object): ...
class Other(object): ...

# Wrong:
class Some(object):
    class Inner(object):
        ...

New in version 0.1.0.

error_template = 'Found nested class: {0}'

code = 431

class MagicNumberViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use magic numbers in your code.

What we call a “magic number”? Well, it is actually any number that appears in your code out of nowhere. Like 42. Or 0.32.

Reasoning:

It is very hard to remember what these numbers actually mean. Why were they used? Should they ever be changed? Or are they eternal like 3.14?

Solution:

Give these numbers a name! Move them to a separate variable, giving more context to the reader. And by moving things into new variables you will trigger other complexity checks.

Example:

# Correct:
price_in_euro = 3.33  # could be changed later
total = get_items_from_cart() * price_in_euro

# Wrong:
total = get_items_from_cart() * 3.33

What are numbers that we exclude from this check? Any numbers that are assigned to a variable, array, dictionary, or keyword arguments inside a function. int numbers that are in range [-10, 10] and some other common numbers, that are defined in MAGIC_NUMBERS_WHITELIST

New in version 0.1.0.

See also

https://en.wikipedia.org/wiki/Magic_number_(programming)

code = 432

error_template = 'Found magic number: {0}'

class StaticMethodViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use @staticmethod decorator.

Reasoning:

Static methods are not required to be inside the class. Because they even do not have access to the current instance.

Solution:

Use instance methods, @classmethod, or functions instead.

New in version 0.1.0.

error_template = 'Found using `@staticmethod`'

code = 433

class BadMagicMethodViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use some magic methods.

Reasoning:

We forbid to use magic methods related to the forbidden language parts. Likewise, we forbid to use del keyword, so we forbid to use all magic methods related to it.

Solution:

Refactor you code to use custom methods instead. It will give more context to your app.

See MAGIC_METHODS_BLACKLIST for the full blacklist of the magic methods.

New in version 0.1.0.

See also

https://www.youtube.com/watch?v=F6u5rhUQ6dU

error_template = 'Found using restricted magic method: {0}'

code = 434

class NestedImportViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have nested imports in functions.

Reasoning:

Usually nested imports are used to fix the import cycle. So, nested imports show that there’s an issue with you design.

Solution:

You don’t need nested imports, you need to refactor your code. Introduce a new module or find another way to do what you want to do. Rethink how your layered architecture should look like.

Example:

# Correct:
from my_module import some_function

def some(): ...

# Wrong:
def some():
    from my_module import some_function

New in version 0.1.0.

See also

https://github.com/seddonym/layer_linter

error_template = 'Found nested import'

code = 435

class RedundantLoopElseViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use else without break in a loop.

We use the same logic for for and while loops.

Reasoning:

When there’s no break keyword in loop’s body it means that else will always be called. This rule will reduce complexity, improve readability, and protect from possible errors.

Solution:

Refactor your else case logic to be inside the loop’s body. Or right after it.

Example:

# Correct:
for letter in 'abc':
    if letter == 'b':
        break
else:
    print('"b" is not found')

for letter in 'abc':
    print(letter)
print('always called')

# Wrong:
for letter in 'abc':
    print(letter)
else:
    print('always called')

New in version 0.3.0.

error_template = 'Found `else` in a loop without `break`'

code = 436

class RedundantFinallyViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use finally in try block without except block.

Reasoning:

This rule will reduce complexity and improve readability.

Solution:

Refactor your try logic. Replace the try-finally statement with a with statement.

Example:

# Correct:
with open("filename") as f:
    f.write(...)

# Wrong:
try:
    f = open("filename")
    f.write(...)
finally:
    f.close()

New in version 0.3.0.

error_template = 'Found `finally` in `try` block without `except`'

code = 437

class ReassigningVariableToItselfViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to assign variable to itself.

Reasoning:

There is no need to do that. Generally it is an indication of some error or just dead code.

Example:

# Correct:
some = some + 1
x_coord, y_coord = y_coord, x_coord

# Wrong:
some = some
x_coord, y_coord = x_coord, y_coord

New in version 0.3.0.

error_template = 'Found reassigning variable to itself'

code = 438

class YieldInsideInitViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use yield inside of __init__ method.

Reasoning:

__init__ should be used to initialize new objects. It shouldn’t yield anything, because it should return None by the convention.

Example:

 # Correct:
class Example(object):
    def __init__(self):
        self._public_items_count = 0

# Wrong:
class Example(object):
    def __init__(self):
        yield 10

New in version 0.3.0.

error_template = 'Found `yield` inside `__init__` method'

code = 439

class ProtectedModuleViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to import protected modules.

Reasoning:

When importing protected modules we break a contract that authors of this module enforce. This way we are not respecting encapsulation and it may break our code at any moment.

Solution:

Do not import anything from protected modules. Respect the encapsulation.

Example:

# Correct:
from some.public.module import FooClass

# Wrong:
import _compat
from some._protected.module import BarClass
from some.module import _protected

New in version 0.3.0.

error_template = 'Found protected module import'

code = 440

class ProtectedAttributeViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use protected attributes and methods.

Reasoning:

When using protected attributes and method we break a contract that authors of this class enforce. This way we are not respecting encapsulation and it may break our code at any moment.

Solution:

Do not use protected attributes and methods. Respect the encapsulation.

Example:

# Correct:
self._protected = 1
cls._hidden_method()
some.public()

# Wrong:
print(some._protected)
instance._hidden()
self.container._internal = 10

Note, that it is possible to use protected attributes with self and cls as base names. We allow this so you can create and use protected attributes and methods inside the class context. This is how protected attributes should be used.

New in version 0.3.0.

error_template = 'Found protected attribute usage: {0}'

code = 441

class LambdaInsideLoopViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to use lambda inside loops.

Reasoning:

It is error-prone to use lambda inside for and while loops due to the famous late-binding.

Solution:

Use regular functions, factory functions, or partial functions. Save yourself from possible confusion.

Example:

# Correct:
for index in range(10):
    some.append(partial_function(index))

# Wrong:
for index in range(10):
    some.append(lambda index=index: index * 10))
    other.append(lambda: index * 10))

New in version 0.5.0.

See also

https://docs.python-guide.org/writing/gotchas/#late-binding-closures

error_template = "Found `lambda` in loop's body"

code = 442

class UnreachableCodeViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have unreachable code.

What is unreachable code? It is some lines of code that can not be executed by python’s interpreter.

This is probably caused by return or raise statements. However, we can not cover 100% of truly unreachable code by this rule. This happens due to the dynamic nature of python. For example, detecting that 1 / some_value would sometimes raise an exception is too complicated and is out of scope of this rule.

Reasoning:

Having dead code in your project is an indicator that you do not care about your code base at all. It dramatically reduces code quality and readability. It also demotivates team members.

Solution:

Delete any unreachable code your have. Or refactor it, if this happens by your mistake.

Example:

# Correct:
def some_function():
    print('This line is reachable, all good')
    return 5

# Wrong:
def some_function():
    return 5
    print('This line is unreachable')

New in version 0.5.0.

error_template = 'Found unreachable code'

code = 443

class StatementHasNoEffectViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have statements that do nothing.

Reasoning:

Statements that just access the value, or expressions used as statements indicate that your code contains dead lines. They just pollute your codebase and do nothing.

Solution:

Refactor your code in case it was a typo or error. Or just delete this code.

Example:

# Correct:
def some_function():
    price = 8 + 2
    return price

# Wrong:
def some_function():
    8 + 2
    print

New in version 0.5.0.

error_template = 'Found statement that has no effect'

code = 444

class MultipleAssignmentsViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have statements that do nothing.

Reasoning:

Multiple assignments on the same line might not do what you think they do. They can also grown pretty long. And you will not notice the rising complexity of your code.

Solution:

Use separate lines for each assignment.

Example:

# Correct:
a = 1
b = 1

# Wrong:
a = b = 1

New in version 0.6.0.

error_template = 'Found multiple assign targets'

code = 445

class IncorrectUnpackingViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have statements that do nothing.

Reasoning:

Having unpacking with side-effects is very dirty. You might get in serious and very hard-to-debug troubles because of this technique. So, do not use it.

Solution:

Use unpacking with only variables, not any other entities.

Example:

# Correct:
first, second = some()

# Wrong:
first, some_dict['alias'] = some()

New in version 0.6.0.

error_template = 'Found incorrect unpacking target'

code = 446

class DuplicateExceptionViolation(node, text=None)

Bases: wemake_python_styleguide.violations.base.ASTViolation

Forbids to have the same exception class in multiple except blocks.

Reasoning:

Having the same exception name in different blocks means that something is not right: since only one branch will work. Other one will always be ignored. So, that is clearly an error.

Solution:

Use unique exception handling rules.

Example:

# Correct:
try:
    ...
except ValueError:
    ...

# Wrong:
try:
    ...
except ValueError:
    ...
except ValueError:
    ...

New in version 0.6.0.

error_template = 'Found duplicate exception: {0}'

code = 447