tox is an environment orchestrator. Use it to define how to setup and execute various tools on your projects. The tool can be:
- a test runner (such as
pytest
), - a linter (e.g.,
flake8
), - a formatter (for example
black
orisort
), - a documentation generator (e.g.,
Sphinx
), - library builder and publisher (e.g.,
build
withtwine
), - or anything else you may need to execute.
First, in a configuration file you need to define what tools you need to run and how to provision a test environment for these. The canonical file for this is the tox.ini
file, let's take a look at an example of this (this needs to live at the root of your project):
Note
You can also generate a tox.ini
file automatically by running tox quickstart
and then answering a few questions.
[tox]
envlist =
format
py310
[testenv:format]
description = install black in a virtual environment and invoke it on the current folder
deps = black==22.3.0
skip_install = true
commands = black .
[testenv:py310]
description = install pytest in a virtual environment and invoke it on the tests folder
deps =
pytest>=7
pytest-sugar
commands = pytest tests {posargs}
The configuration is split into two type of configuration: core settings are hosted under the tox
section and per run environment settings hosted under testenv:<env_name>
. Under the core section we define that this project has two run environments named format
and py310
respectively (we use the envlist
configuration key to do so).
Then we define separately what should the formatting environment (testenv:format
section) and the test environment (testenv:py310
section). For example to format the project we:
- add a description (visible when you type
tox list
into the command line), - we define that it requires the
black
PyPI dependency with version22.3.0
, - the black tool does not need the project we are testing to be installed into the test environment so we disable this default behaviour via the
skip_install
configuration, - and we define that the tool should be invoked as we'd type
black .
into the command line.
For testing the project we use the py310
environment, for which we:
- define a text description of the environment,
- specify that requires
pytest
7
or later together with thepytest-sugar
project, - and that the tool should be invoked via the
pytest tests
CLI command.
{posargs}
is a place holder part for the CLI command that allows us to pass additional flags to the pytest invocation, for example if we'd want to run pytest tests -v
as a one off, instead of tox run -e py310
we'd type tox run -e py310 -- -v
. The --
delimits flags for the tox tool and what should be forwarded to the tool within.
tox, by default, always creates a fresh virtual environment for every run environment. The Python version to use for a given environment can be controlled via the base_python
configuration, however if not set will try to use the environment name to determine something sensible: if the name is in the format of pyxy
then tox will create an environment with CPython with version x.y
(for example py310
means CPython 3.10
). If the name does not match this pattern it will use a virtual environment with the same Python version as the one tox is installed into (this is the case for format
).
tox environments are reused between runs, so while the first tox run -e py310
will take a while as tox needs to create a virtual environment and install pytest
and pytest-sugar
in it, subsequent runs only need to reinstall your project, as long as the environments dependency list does not change.
Almost every step and aspect of virtual environments and command execution can be customized. You'll find an exhaustive list of configuration flags (together with what it does and detailed explanation of what values are accepted) at our configuration page <configuration>
.
Below is a graphical representation of the tox states and transition pathways between them:
The primary tox states are:
- Configuration: load tox configuration files (such as
tox.ini
,pyproject.toml
andtoxfile.py
) and merge it with options from the command line plus the operating system environment variables. - Environment: for each selected tox environment (e.g.
py310
,format
) do:- Creation: create a fresh environment; by default
virtualenv
is used, but configurable viarunner
. For virtualenv tox will use the virtualenv discovery logic where the python specification is defined by the tox environmentsbase_python
(if not set will default to the environments name). This is created at first run only to be re-used at subsequent runs. If certain aspects of the project change (python version, dependencies removed, etc.), a re-creation of the environment is automatically triggered. To force the recreation tox can be invoked with therecreate
flag (-r
). - Install dependencies (optional): install the environment dependencies specified inside the
deps
configuration section, and then the earlier packaged source distribution. By defaultpip
is used to install packages, however one can customize this viainstall_command
. Notepip
will not update project dependencies (specified either in theinstall_requires
or theextras
section of thesetup.py
) if any version already exists in the virtual environment; therefore we recommend to recreate your environments whenever your project dependencies change. - Packaging (optional): create a distribution of the current project.
- Build: If the tox environment has a package configured tox will build a package from the current source tree. If multiple tox environments are run and the package built are compatible in between them then it will be reused. This is to ensure that we build the package as rare as needed. By default for Python a source distribution is built as defined via the
pyproject.toml
style build (see PEP-517 and PEP-518). - Install the package dependencies. If this has not changed since the last run this step will be skipped.
- Install the package. This operation will force reinstall the package without its dependencies.
- Build: If the tox environment has a package configured tox will build a package from the current source tree. If multiple tox environments are run and the package built are compatible in between them then it will be reused. This is to ensure that we build the package as rare as needed. By default for Python a source distribution is built as defined via the
- Commands: run the specified commands in the specified order. Whenever the exit code of any of them is not zero, stop and mark the environment failed. When you start a command with a dash character, the exit code will be ignored.
- Creation: create a fresh environment; by default
Report print out a report of outcomes for each tox environment:
____________________ summary ____________________ py37: commands succeeded ERROR: py38: commands failed
Only if all environments ran successfully tox will return exit code
0
(success). In this case you'll also see the messagecongratulations :)
.
tox will take care of environment variable isolation for you. That means it will remove system environment variables not specified via passenv
. Furthermore, it will also alter the PATH
variable so that your commands resolve within the current active tox environment. In general, all executables outside of the tox environment are available in commands
, but external commands need to be explicitly allowed via the allowlist_externals
configuration.
- automation of tedious Python related test activities
- test your Python package against many interpreter and dependency configurations
- automatic customizable (re)creation of
virtualenv
test environments - installs your project into each virtual environment
- test-tool agnostic: runs pytest, nose or unittest in a uniform manner
- automatic customizable (re)creation of
plugin system
to modify tox execution with simple hooks.- uses
pip
andvirtualenv
by default. Support for plugins replacing it with their own. - cross-Python compatible: tox requires CPython 3.7 and higher, but it can create environments 2.7 or later
- cross-platform: Windows, macOS and Unix style environments
- full interoperability with devpi: is integrated with and is used for testing in the
devpi
system, a versatile PyPI index server and release managing tool - driven by a simple (but flexible to allow expressing more complicated variants) ini-style config file
- documented examples and configuration
- concise reporting about tool invocations and configuration errors
- supports using different / multiple PyPI index servers
tox has influenced several other projects in the Python test automation space. If tox doesn't quite fit your needs or you want to do more research, we recommend taking a look at these projects:
- nox is a project similar in spirit to tox but different in approach. The primary key difference is that it uses Python scripts instead of a configuration file. It might be useful if you find tox configuration too limiting but aren't looking to move to something as general-purpose as
Invoke
ormake
. Please note that tox will support defining configuration in a Python file soon, too. - Invoke is a general-purpose task execution library, similar to Make. Invoke is far more general-purpose than tox but it does not contain the Python testing-specific features that tox specializes in.
In case the installed tox version does not satisfy either the min_version
or the requires
, tox will automatically create a virtual environment under provision_tox_env
name that satisfies those constraints and delegate all calls to this meta environment. This should allow satisfying constraints on your tox environment automatically, given you have at least version 3.8.0
of tox.
For example given:
[tox]
min_version = 4
requires = tox-docker>=1
if the user runs it with tox 3.8
or later the installed tox application will automatically ensure that both the minimum version and requires constraints are satisfied, by creating a virtual environment under .tox
folder, and then installing into it tox>=4
and tox-docker>=1
. Afterwards all tox invocations are forwarded to the tox installed inside .tox\.tox
folder (referred to as meta-tox or auto-provisioned tox).
This allows tox to automatically setup itself with all its plugins for the current project. If the host tox satisfies the constraints expressed with the requires
and min_version
no such provisioning is done (to avoid setup cost and indirection when it's not explicitly needed).
This section details information that you'll use most often in short form.
- Each tox subcommand has a 1 (or 2) letter shortcut form too, e.g.
tox run
can also be written astox r
ortox config
can be shortened totox c
. - To run all tox environments defined in the
env_list
run tox without any flags:tox
. - To run a single tox environment use the
-e
flag for therun
sub-command as intox run -e py310
. - To run two or more tox environment pass comma separated values, e.g.
tox run -e format,py310
. The run command will run the tox environments sequentially, one at a time, in the specified order. - To run two or more tox environment in parallel use the
parallel
sub-command , e.g.tox parallel -e py39,py310
. The--parallel
flag for this sub-command controls the degree of parallelism. - To view the configuration value for a given environment and a given configuration key use the config sub-command with the
-k
flag to filter for targeted configuration values:tox config -e py310 -k pass_env
. - tox tries to automatically detect changes to your project dependencies and force a recreation when needed. Unfortunately the detection is not always accurate, and it also won't detect changes on the PyPI index server. You can force a fresh start for the tox environments by passing the
-r
flag to your run command. Whenever you see something that should work but fails with some esoteric error it's recommended to use this flag to make sure you don't have a stale Python environment; e.g.tox run -e py310 -r
would clean the run environment and recreate it from scratch.
Every tox environment has its own configuration section (e.g. in case of
tox.ini
configuration method thepy310
tox environments configuration is read from thetestenv:py310
section). If the section is missing or does not contain that configuration value, it will fall back to the section defined by thebase
configuration (fortox.ini
this is thetestenv
section). For example:[testenv] commands = pytest tests [testenv:test] description = run the test suite with pytest
Here the environment description for
test
is taken fromtestenv:test
. Ascommands
is not specified, the value defined under thetestenv
section will be used. If the base environment is also missing a configuration value then the configuration default will be used (e.g. in case of thepass_env
configuration here).- To change the current working directory for the commands run use
change_dir
(note this will make the change for all install commands too - watch out if you have relative paths in your project dependencies). - Environment variables:
- To view environment variables set and passed down use
tox c -e py310 -k set_env pass_env
. - To pass through additional environment variables use
pass_env
. - To set environment variables use
set_env
.
- To view environment variables set and passed down use
- Setup operation can be configured via the
commands_pre
, while teardown commands via thecommands_post
. Configurations may be set conditionally within the
tox.ini
file. If a line starts with an environment name or names, separated by a comma, followed by:
the configuration will only be used if the environment name(s) matches the executed tox environment. For example:[testenv] deps = pip format: black py310,py39: pytest
Here pip will be always installed as the configuration value is not conditional. black is only used for the
format
environment, whilepytest
is only installed for thepy310
andpy39
environments.
tox
allows running environments in parallel mode via the parallel
sub-command:
- After the packaging phase completes tox will run the tox environments in parallel processes (multi-thread based).
- the
--parallel
flag takes an argument specifying the degree of parallelization, defaulting toauto
:all
to run all invoked environments in parallel,auto
to limit it to CPU count,- or pass an integer to set that limit.
- Parallel mode displays a progress spinner while running tox environments in parallel, and reports outcome of these as soon as they have been completed with a human readable duration timing attached. This spinner can be disabled via the
--parallel-no-spinner
flag. - Parallel mode by default shows output only of failed environments and ones marked as
parallel_show_output
=True
. There's now a concept of dependency between environments (specified via
depends
), tox will re-order the environment list to be run to satisfy these dependencies, also for sequential runs. Furthermore, in parallel mode, tox will only schedule a tox environment to run once all of its dependencies have finished (independent of their outcome).Warning
depends
does not pull in dependencies into the run target, for example if you selectpy310,py39,coverage
via the-e
tox will only run those three (even ifcoverage
may specify asdepends
other targets too -such aspy310, py39, py38, py37
).--parallel-live
/-o
allows showing the live output of the standard output and error, also turns off reporting as described above.- Note: parallel evaluation disables standard input. Use non parallel invocation if you need standard input.
Example final output:
$ tox -e py310,py39,coverage -p all
✔ OK py39 in 9.533 seconds
✔ OK py310 in 9.96 seconds
✔ OK coverage in 2.0 seconds
___________________________ summary ______________________________________________________
py310: commands succeeded
py39: commands succeeded
coverage: commands succeeded
congratulations :)
Example progress bar, showing a rotating spinner, the number of environments running and their list (limited up to 120 characters):
⠹ [2] py310 | py39
tox always builds projects in a PEP-518 compatible virtual environment and communicates with the build backend according to the interface defined in PEP-517 and PEP-660. To define package build dependencies and specify the build backend to use create a pyproject.toml
at the root of the project. For example to use hatch:
[build-system]
build-backend = "hatchling.build"
requires = ["hatchling>=0.22", "hatch-vcs>=0.2"]
By default tox will create and install a source distribution. You can configure to build a wheel instead by setting the package
configuration to wheel
. Wheels are much faster to install than source distributions.
To query the projects dependencies tox will use a virtual environment whose name is defined under the package_env
configuration (by default .pkg
). The virtual environment used for building the package depends on the artifact built:
- for source distribution the
package_env
, - for wheels the name defined under
wheel_build_env
(this depends on the Python version defined by the target tox environment underbase_python
, if the environment targets CPython 3.10 it will be.pkg-cpython310
or for PyPy 3.9 it will be.pkg-pypy39
).
For pure Python projects (non C-Extension ones) it's recommended to set wheel_build_env
to the same as the package_env
. This way you'll build the wheel once and install the same wheel for all tox environments.
tox supports these features that 90 percent of the time you'll not need, but are very useful the other ten percent.
If you have a large matrix of dependencies, python versions and/or environments you can use a generative env_list
and conditional settings to express that in a concise form:
[tox]
env_list = py{311,310,39}-django{41,40}-{sqlite,mysql}
[testenv]
deps =
django41: Django>=4.1,<4.2
django40: Django>=4.0,<4.1
# use PyMySQL if factors "py311" and "mysql" are present in env name
py311-mysql: PyMySQL
# use urllib3 if any of "py311" or "py310" are present in env name
py311,py310: urllib3
# mocking sqlite on 3.11 and 3.10 if factor "sqlite" is present
py{311,310}-sqlite: mock
This will generate the following tox environments:
> tox l
default environments:
py311-django41-sqlite -> [no description]
py311-django41-mysql -> [no description]
py311-django40-sqlite -> [no description]
py311-django40-mysql -> [no description]
py310-django41-sqlite -> [no description]
py310-django41-mysql -> [no description]
py310-django40-sqlite -> [no description]
py310-django40-mysql -> [no description]
py39-django41-sqlite -> [no description]
py39-django41-mysql -> [no description]
py39-django40-sqlite -> [no description]
py39-django40-mysql -> [no description]
Suppose you have some binary packages, and need to run tests both in 32 and 64 bits. You also want an environment to create your virtual env for the developers.
[testenv]
base_python =
py311-x86: python3.11-32
py311-x64: python3.11-64
commands = pytest
[testenv:py311-{x86,x64}-venv]
envdir =
x86: .venv-x86
x64: .venv-x64
> tox l
default environments:
py -> [no description]
additional environments:
py310-black -> [no description]
py310-lint -> [no description]
py311-black -> [no description]
py311-lint -> [no description]