Skip to main content

Add the mutation testing feature to pytest

Project description


Mutagen is a plugin to pytest that makes it easy to do mutation testing. Mutation testing is a method of testing your tests. Mutagen helps you to define "mutant" versions of your code---code which is intentionally buggy---then you run your test suite on these mutants and verify that your tests actually catch the bugs. Mutation testing helps you to gauge test coverage and verify that your tests are good enough to exercise interesting behaviors in your code.

For Property-Based Testing

If you are a user of a property-based testing framework such as Hypothesis, mutation testing can also be used to test your input generators. It is relatively easy to write a generator that cannot generate a certain kind of input. Mutation testing can be used to find those gaps.


python3 -m pip install pytest-mutagen


Python import

import pytest_mutagen as mg

Declare a mutant

  • Mutant function
    To mutate a whole function you have to write the new version of the function, decorated with @mg.mutant_of(function_qual_name, mutant_name, file (optional), description (optional)). If the mutations affect an object (function or class) you have to be sure that this object exists in the __globals__ symbols table of the mutant functions. For this purpose you can simply write from [your_module] import [target_object] in the mutation file. Example:

     def  inc(x):
     	return x + 1
     @mg.mutant_of("inc", "INC_OBO", description="Increment is off by one.")
     def  inc_mut(x):
     	return x + 2
  • Mutant expression
    If you don't want to change the whole function but only one line, you must decorate the function with @mg.has_mutant(mutant_name, file (optional), description (optional)). Then you have two ways to do it:

    • By replacing the expression by the mg.mut(mutant_name, normal_expression, mutant_expression) function, using lambda expressions.
      Example: mg.mut("FLIP_LT", lambda: a < b, lambda: b < a)

    • Using the mg.not_mutant(mutant_name) function combined with an if statement.
      Example: k = inc(k) if mg.not_mutant("INC_OBO2") else inc(k) + 1

Mutating a class method

In fact the @mutant_of decorator doesn't require the function name but its fully qualified name. It does not change anything for top-level functions but in the case of a class method you need to write the dotted path leading to the object from the module top-level. Example:

class Foo:
	def bar(self):

	def static_bar():

@mg.mutant_of("", "")
def bar_mut(self):

@mg.mutant_of("Foo.static_bar", "")
def static_bar_mut():

Global functioning

Mutagen collects all declared mutants, stored per file names. Then it looks through all tests collected by pytest and apply the mutants to the matching files. This is handled by the optional file parameter in @has_mutant and @mutant_of which can be a file name or a list of file names where you want your mutant to be applied. You can set it to APPLY_TO_ALL (constant string declared in mutagen) if you want it to be applied to all collected files. By default, file is:

  • APPLY_TO_ALL for @has_mutant
  • the current file name for @mutant_of (the one where it is written)

Therefore you can either:

  • write your mutations and specify for each one where you want it to be applied (use the function mg.link_to_file(filename) at the beginning of your file to link the current file to the specified filename)
  • or create a file where you import all test files you want (from import *), write your mutant_of with no file specified and run pytest on

Run the tests

python3 -m pytest --mutate

Quick run

The --quick-mut option will stop each mutant after its first failed test. If not specified each mutant will run the whole test suite

Cache use

Mutagen stores in the pytest cache the functions that failed during the last run, for each mutant. For the next runs it will try these functions first, in order to find failures more quickly. If you don't need this feature you can simply use the --cache-clear option that will clear the cache before running the tests.

Run only the mutations

If you don't want to run the original test suite but only the mutations you can use the pytest option --collect-only

Selective run of mutants

The --select option expects a comma-separated list of mutants (no spaces) and will run these ones exclusively. Example:

python3 -m pytest --mutate --select INC_OBO,FLIP_LT

Mutagen stats

The --mutagen-stats option adds a section to the terminal summary, which displays the number of tests that caught each mutant.

Add trivial mutations

To find holes in a test suite with mutagen, we often try trivial mutations on some functions (like replacing them with pass) to see whether a lot of tests catch them or not. For this purpose the trivial_mutations(functions, obj=None, file=APPLY_TO_ALL) function with a list of functions as input adds all mutants corresponding to replacing them by an empty function. There are two ways to use it:

from module import sort, invert, ExampleClass

# With a list of top-level functions
mg.trivial_mutations([sort, invert])

# With a list of method names and the corresponding object
mg.trivial_mutations(["sort", "clear"], ExampleClass)

This is equivalent to doing this:

from module import sort, invert, ExampleClass


@mg.mutant_of("sort", "SORT_NOTHING")
def sort_mut(*args, **kwargs):

@mg.mutant_of("invert", "INVERT_NOTHING")
def invert_mut(*args, **kwargs):

@mg.mutant_of("ExampleClass.sort", "EXAMPLECLASS.SORT_NOTHING")
def sort_mut(*args, **kwargs):

@mg.mutant_of("ExampleClass.clear", "EXAMPLECLASS.CLEAR_NOTHING")
def clear_mut(*args, **kwargs):

trivial_mutations has an optional file parameter to specify the test file where the mutations should be applied, which is by default set to APPLY_TO_ALL.

The function trivial_mutations_all(object, file=APPLY_TO_ALL) applies this process to each method of the class (or list of classes) given as a parameter. Example:

from module import ExampleClass



You can find some examples in the examples folder

  • The file is a very simple example of the use of mutagen to test a merge sort function
  • The file implements the Binary Search Tree data structure, and the test suite and mutations from How to specify it! (John Hughes, 2019)
  • The subfolder separate_files is an example of the separation between the source file, the test file and the mutation file

The scripts show how to run these tests

Automatic tool

Writing mutations by hand can be very long and we are aware that it can discourage a lot of programmers from using pytest-mutagen, that is initially a manual mutation-testing tool. To fix this problem while keeping the ability to manually edit the mutants we've added the possibility to interactively generate a bunch of mutants following a set of rules.


python3 -m pytest_mutagen [-h] [-o OUTPUT_PATH] [-m MODULE_PATH] input_path

This command will browse the provided input_path (that can be a file or a directory) and interactively propose several mutants. You can accept them by pressing ENTER and refuse them by typing 'n' then ENTER. The purpose of this is to avoid false positives and equivalent mutants, that are among the main problems of mutation testing. Finally all accepted mutants are written in the mutagen syntax (ready to be use with pytest --mutate) in or the file/directory specified with the -o command-line option. For more details on its use you can use python3 -m pytest_mutagen --help.


  • Integers are incremented
  • Operators are switched to a different (but close) operator
  • In assignments, the right value is replaced with None
  • The return statement is removed
  • The condition of if statements are replaced with not (condition)

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pytest-mutagen-1.3.tar.gz (18.1 kB view hashes)

Uploaded Source

Built Distribution

pytest_mutagen-1.3-py3-none-any.whl (17.0 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page