# Complexity¶

These checks finds flaws in your application design.

We try to stick to “the magical 7 ± 2 number” when counting things. https://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_or_Minus_Two

That’s how many objects we can keep in our memory at a time. We try hard not to exceed the memory capacity limit.

Note

Simple is better than complex. Complex is better than complicated.

## Summary¶

 JonesScoreViolation Forbids to have modules with complex lines. TooManyImportsViolation Forbids to have modules with too many imports. TooManyModuleMembersViolation Forbids to have many classes and functions in a single module. TooManyLocalsViolation Forbids to have too many local variables in the unit of code. TooManyArgumentsViolation Forbids to have too many arguments for a function or method. TooManyReturnsViolation Forbids placing too many return statements into the function. TooManyExpressionsViolation Forbids putting too many expressions in a unit of code. TooManyMethodsViolation Forbids to have many methods in a single class. TooDeepNestingViolation Forbids nesting blocks too deep. LineComplexityViolation Forbids to have complex lines. TooManyConditionsViolation Forbids to have conditions with too many logical operators. TooManyElifsViolation Forbids to use many elif branches. TooManyForsInComprehensionViolation Forbids to have too many for statement within a comprehension. TooManyBaseClassesViolation Restrict the maximum number of base classes.

## Module complexity¶

class JonesScoreViolation(node=None, text=None)[source]

Forbids to have modules with complex lines.

We are using Jones Complexity algorithm to count module’s score. See LineComplexityViolation for details of per-line-complexity. How it is done: we count complexity per line, then measuring the median complexity across the lines in the whole module.

Reasoning:
Having complex modules will decrease your code maintainability.
Solution:
Refactor the module contents.
Configuration:
This rule is configurable with --max-jones-score. Default: 12

New in version 0.1.0.

Note

Returns Z200 as error code

error_template = 'Found module with high Jones Complexity score: {0}'

Error message shown to the user.

class TooManyImportsViolation(node=None, text=None)[source]

Forbids to have modules with too many imports.

Namespaces are one honking great idea – let’s do more of those!

Reasoning:

Having too many imports without prefixes is quite expensive. You have to memorize all the source locations of the imports. And sometimes it is hard to remember what kind of functions and classes are already injected into your context.

It is also a questionable design if a single module has a lot of imports. Why a single module has so many dependencies? So, it becomes too coupled.

Solution:

Refactor the imports to import a common namespace. Something like from package import module and then use it like module.function().

Or refactor your code and split the complex module into several ones.

We do not make any differences between import and from ... import ....

Configuration:
This rule is configurable with --max-imports. Default: 12

New in version 0.1.0.

Note

Returns Z201 as error code

error_template = 'Found module with too many imports: {0}'

Error message shown to the user.

class TooManyModuleMembersViolation(node=None, text=None)[source]

Forbids to have many classes and functions in a single module.

Reasoning:
Having many classes and functions in a single module is a bad thing. Soon it will be hard to read through this code and understand it.
Solution:
It is better to split this module into several modules or a package.

We do not make any differences between classes and functions in this check. They are treated as the same unit of logic. We also do not care about functions and classes being public or not. However, methods are counted separately on a per-class basis.

Configuration:
This rule is configurable with --max-module-members. Default: 7

New in version 0.1.0.

Note

Returns Z202 as error code

error_template = 'Found too many module members: {0}'

Error message shown to the user.

class TooManyBaseClassesViolation(node, text=None)[source]

Restrict the maximum number of base classes.

Reasoning:
It is almost never possible to navigate to the desired method of a parent class when you need it with multiple mixins. It is hard to understand mro and super calls. Do not overuse this technique.
Solution:
Restrict the number of base classes.

Example:

# Correct:
class SomeClassName(FirstParentClass, SecondParentClass): ...

# Wrong:
class SomeClassName(
FirstParentClass,
SecondParentClass,
ThirdParentClass,
CustomClass,
): ...
Configuration:
This rule is configurable with --max-base-classes. Default: 3

New in version 0.3.0.

Note

Returns Z225 as error code

error_template = 'Too many number of base classes: {0}'

Error message shown to the user

## Function and class complexity¶

class TooManyLocalsViolation(node, text=None)[source]

Forbids to have too many local variables in the unit of code.

Reasoning:
Having too many variables in a single function is bad thing. Soon, you will find troubles to understand what this variable means. It will also become hard to name new variables.
Solution:
If you have too many variables in a function, you have to refactor it.

What counts as a local variable? We only count variable as local in the following case: it is assigned inside the function body. We do not count variables defined inside comprehensions as local variables, since it is impossible to use them outside of the comprehension.

Example:

def first_function(param):
first_var = 1

def second_function(argument):
second_var = 1
argument = int(argument)
third_var, _ = some_call()

In this example we will count as locals only several variables:

1. first_var, because it is assigned inside the function’s body
2. second_var, because it is assigned inside the function’s body
3. argument, because it is reassigned inside the function’s body
4. third_var, because it is assigned inside the function’s body

Please, note that _ is a special case. It is not counted as a local variable. Since by design it means: do not count me as a real variable.

Configuration:
This rule is configurable with --max-local-variables. Default: 5

New in version 0.1.0.

Note

Returns Z210 as error code

error_template = 'Found too many local variables: {0}'

Error message shown to the user.

class TooManyArgumentsViolation(node, text=None)[source]

Forbids to have too many arguments for a function or method.

Reasoning:
This is an indicator of a bad design. When function requires many arguments it shows that it is required to refactor this piece of code. It also indicates that function does too many things at once.
Solution:
Split function into several functions. Then it will be easier to use them.
Configuration:
This rule is configurable with --max-arguments. Default: 5

New in version 0.1.0.

Note

Returns Z211 as error code

error_template = 'Found too many arguments: {0}'

Error message shown to the user.

class TooManyReturnsViolation(node, text=None)[source]

Forbids placing too many return statements into the function.

Reasoning:
When there are too many return keywords, functions are hard to test. They are also hard to read and hard to change and keep everything inside your head at once.
Solution:
Configuration:
This rule is configurable with --max-returns. Default: 5

New in version 0.1.0.

Note

Returns Z212 as error code

error_template = 'Found too many return statements: {0}'

Error message shown to the user.

class TooManyExpressionsViolation(node, text=None)[source]

Forbids putting too many expressions in a unit of code.

Reasoning:
When there are too many expressions it means that this specific function does too many things at once. It has too much logic.
Solution:
Split function into several functions, refactor your API.
Configuration:
This rule is configurable with --max-expressions. Default: 9

New in version 0.1.0.

Note

Returns Z213 as error code

error_template = 'Found too many expressions: {0}'

Error message shown to the user.

class TooManyMethodsViolation(node, text=None)[source]

Forbids to have many methods in a single class.

Reasoning:
Having too many methods might lead to the “God object”. This kind of objects can handle everything. So, in the end your code becomes too hard to maintain and test.
Solution:
What to do if you have too many methods in a single class? Split this class into several classes. Then use composition or inheritance to refactor your code. This will protect you from “God object” anti-pattern.

We do not make any difference between instance and class methods. We also do not care about functions and classes being public or not. We also do not count inherited methods from parents. This rule does not count attributes of a class.

Configuration:
This rule is configurable with --max-methods. Default: 7

New in version 0.1.0.

Note

Returns Z214 as error code

error_template = 'Found too many methods: {0}'

Error message shown to the user.

## Structures complexity¶

class TooDeepNestingViolation(node, text=None)[source]

Forbids nesting blocks too deep.

Reasoning:
If nesting is too deep that indicates usage of a complex logic and language constructions. This means that our design is not suited to handle such construction.
Solution:
We need to refactor our complex construction into simpler ones. We can use new functions or different constructions.
Configuration:
This rule is configurable with --max-offset-blocks. Default: 5

New in version 0.1.0.

Note

Returns Z220 as error code

error_template = 'Found too deep nesting: {0}'

Error message shown to the user.

class LineComplexityViolation(node, text=None)[source]

Forbids to have complex lines.

We are using Jones Complexity algorithm to count complexity. What is Jones Complexity? It is a simple yet powerful method to count the number of ast nodes per line. If the complexity of a single line is higher than a threshold, then an error is raised.

What nodes do we count? All except the following:

1. modules
2. function and classes, since they are checked differently
3. type annotations, since they do not increase complexity
Reasoning:
Having a complex line indicates that you somehow managed to put too much logic inside a single line. At some point in time you will no longer be able to understand what this line means and what it does.
Solution:
Split a single line into several lines: by creating new variables, statements or functions. Note, this might trigger new complexity issues. With this technique a single new node in a line might trigger a complex refactoring process including several modules.
Configuration:
This rule is configurable with --max-line-complexity. Default: 14

New in version 0.1.0.

Note

Returns Z221 as error code

error_template = 'Found line with high Jones Complexity: {0}'

Error message shown to the user.

class TooManyConditionsViolation(node, text=None)[source]

Forbids to have conditions with too many logical operators.

Reasoning:
When reading through the complex conditions you will fail to understand all the possible branches. And you will end up putting debug breakpoint on this line just to figure out how it works.
Solution:
We can reduce the complexity of a single if by doing two things: creating new variables or creating nested if statements. Both of these actions will trigger other complexity checks.

We only check if and while nodes for this type of complexity. We check if nodes inside list comprehensions and ternary expressions.

We count and and or keywords as conditions.

Example:

# The next line has 2 conditions:
if x_coord > 1 and x_coord < 10: ...
Configuration:
This rule is configurable with --max-conditions. Default: 4

New in version 0.1.0.

Note

Returns Z222 as error code

error_template = 'Found a condition with too much logic: {0}'

Error message shown to the user.

class TooManyElifsViolation(node, text=None)[source]

Forbids to use many elif branches.

Reasoning:
This rule is specifically important, because many elif branches indicate a complex flow in your design: you are reimplementing switch in python.
Solution:
There are different design patters to use instead. For example, you can use some interface that just call a specific method without if.
Configuration:
This rule is configurable with --max-elifs. Default: 3

New in version 0.1.0.

Note

Returns Z223 as error code

error_template = 'Found too many elif branches: {0}'

Error message shown to the user.

class TooManyForsInComprehensionViolation(node, text=None)[source]

Forbids to have too many for statement within a comprehension.

Reasoning:
When reading through the complex comprehension you will fail to understand it.
Solution:
We can reduce the complexity of a comprehension by reducing the amount of for statements. Refactor your code to use several for loops, comprehensions, or different functions.

Example:

# Wrong:
ast_nodes = [
target
for assignment in top_level_assigns
for target in assignment.targets
for _ in range(10)
]

New in version 0.3.0.

Note

Returns Z224 as error code

error_template = 'Found a comprehension with too many for statements'

Error message shown to the user.