Developer’s Guide


Welcome to the xonsh developer’s guide! This is a place for developers to place information that does not belong in the user’s guide or the library reference but is useful or necessary for the next people that come along to develop xonsh.


All code changes must go through the pull request review procedure.

Making Your First Change

First, install xonsh from source and open a xonsh shell in your favorite terminal application. See installation instructions for details, but it is recommended to do an ‘editable’ install via `pip’

$ pip install -e .

Next, make a trivial change (e.g. print("hello!") in

Finally, run the following commands. You should see the effects of your change (e.g. hello!):

$ xonsh

The xonsh build process collapses all Python source files into a single file. When xonsh is started with a falsy value for $XONSH_DEBUG, it imports Python modules straight from, which decreases startup times by eliminating the cost of runtime imports. But setting $ $XONSH_DEBUG=1 will suppress amalgamated imports. Reloading the xonsh shell ($ xonsh) won’t simply import the stale file that doesn’t contain your new change, but will instead import the unamalgamated source code which does contain your change. You can now load every subsequent change by reloading xonsh, and if your code changes don’t seem to have any effect, make sure you check $XONSH_DEBUG first!


Pull requests will often have CHANGELOG entries associated with. However, to avoid excessive merge conflicts, please follow the following procedure:

  1. Go into the news/ directory,

  2. Copy the TEMPLATE.rst file to another file in the news/ directory. We suggest using the branchname:

    $ cp TEMPLATE.rst branch.rst
  3. Add your entries as a bullet pointed lists in your branch.rst file in the appropriate category. It is OK to leave the None entries for later use.

  4. Commit your branch.rst.

Feel free to update this file whenever you want! Please don’t use someone else’s file name. All of the files in this news/ directory will be merged automatically at release time. The None entries will be automatically filtered out too!

Style Guide

xonsh is a pure Python project, and so we use PEP8 (with some additions) to ensure consistency throughout the code base.

Rules to Write By

It is important to refer to things and concepts by their most specific name. When writing xonsh code or documentation please use technical terms appropriately. The following rules help provide needed clarity.


  • User-facing APIs should be as generic and robust as possible.

  • Tests belong in the top-level tests directory.

  • Documentation belongs in the top-level docs directory.


  • Code must have associated tests and adequate documentation.

  • User-interaction code (such as the Shell class) is hard to test. Mechanism to test such constructs should be developed over time.

  • Have extreme empathy for your users.

  • Be selfish. Since you will be writing tests you will be your first user.

Python Style Guide

xonsh uses PEP8 for all Python code. The following rules apply where PEP8 is open to interpretation.

  • Use absolute imports (import rather than explicit relative imports (import .tools). Implicit relative imports (import tools) are never allowed.

  • Use "double quotes" for string literals, and """triple double quotes""" for docstrings.

  • We use sphinx with the numpydoc extension to autogenerate API documentation. Follow the numpydoc standard for docstrings.

  • Simple functions should have simple docstrings.

  • Lines should be at most 80 characters long. The 72 and 79 character recommendations from PEP8 are not required here.

  • All Python code should be compliant with Python 3.5+.

  • Tests should be written with pytest using a procedural style. Do not use unittest directly or write tests in an object-oriented style.

  • Test generators make more dots and the dots must flow!

You can easily check for style issues, including some outright bugs such as misspelled variable names, using flake8. If you’re using Anaconda you’ll need to run “conda install flake8” once. You can easily run flake8 on the edited files in your uncommitted git change:

$ git status -s | awk '/\.py$$/ { print $2 }' | xargs flake8

If you want to lint the entire code base run:

$ flake8

We also use black for formatting the code base (which includes running in our tests):

$ black --check xonsh/ xontrib/

To add this as a git pre-commit hook:

$ pre-commit install


Xonsh source code may be amalgamated into a single file ( to speed up imports. The way the code amalgamater works is that other modules that are in the same package (and amalgamated) should be imported with:

from pkg.x import a, c, d

This is because the amalgamater puts all such modules in the same globals(), which is effectively what the from-imports do. For example, xonsh.ast and xonsh.execer are both in the same package (xonsh). Thus they should use the above from from-import syntax.

Alternatively, for modules outside of the current package (or modules that are not amalgamated) the import statement should be either import pkg.x or import pkg.x as name. This is because these are the only cases where the amalgamater is able to automatically insert lazy imports in way that is guaranteed to be safe. This is due to the ambiguity that from pkg.x import name may import a variable that cannot be lazily constructed or may import a module. So the simple rules to follow are that:

  1. Import objects from modules in the same package directly in using from-import,

  2. Import objects from modules outside of the package via a direct import or import-as statement.

How to Test


If you want to run your “work in progress version” without installing and in a fresh environment you can use Docker. If Docker is installed you just have to run this:

$ python

This will build and run the current state of the repository in an isolated container (it may take a while the first time you run it). There are two additional arguments you can pass this script.

  • The version of python

  • the version of prompt_toolkit


$ python 3.4 0.57

Ensure your cwd is the root directory of the project (i.e., the one containing the .git directory).


Prep your environment for running the tests:

$ pip install -r requirements/tests.txt

Running the Tests - Basic

Run all the tests using pytest:

$ pytest -q

Use “-q” to keep pytest from outputting a bunch of info for every test.

Running the Tests - Advanced

To perform all unit tests:

$ pytest

If you want to run specific tests you can specify the test names to execute. For example to run test_aliases:

$ pytest

Note that you can pass multiple test names in the above examples:

$ pytest

Writing the Tests - Advanced

(refer to pytest documentation)

With the Pytest framework you can use bare assert statements on anything you’re trying to test, note that the name of the test function has to be prefixed with test_:

def test_whatever():
    assert is_true_or_false

The in tests directory defines fixtures for mocking various parts of xonsh for more test isolation. For a list of the various fixtures:

$ pytest --fixtures

when writing tests it’s best to use pytest features i.e. parametrization:

@pytest.mark.parametrize('env', [test_env1, test_env2])
def test_one(env, xession):
    # update the environment variables instead of setting the attribute
    # which could result in leaks to other tests.
    # each run will have the same set of default env variables set.

this will run the test two times each time with the respective test_env. This can be done with a for loop too but the test will run only once for the different test cases and you get less isolation.

With that in mind, each test should have the least assert statements, preferably one.

At the moment, xonsh doesn’t support any pytest plugins.

Happy Testing!

How to Document

Documentation takes many forms. This will guide you through the steps of successful documentation.


No matter what language you are writing in, you should always have documentation strings along with you code. This is so important that it is part of the style guide. When writing in Python, your docstrings should be in reStructured Text using the numpydoc format.

Auto-Documentation Hooks

The docstrings that you have written will automatically be connected to the website, once the appropriate hooks have been setup. At this stage, all documentation lives within xonsh’s top-level docs directory. We uses the sphinx tool to manage and generate the documentation, which you can learn about from the sphinx website. If you want to generate the documentation, first xonsh itself must be installed and then you may run the following command from the docs dir:

~/xonsh/docs $ make html

For each new module, you will have to supply the appropriate hooks. This should be done the first time that the module appears in a pull request. From here, call the new module mymod. The following explains how to add hooks.

Python Hooks

Python API documentation is generated for the entries in docs/api.rst. sphinx-autosummary is used to generate documentation for the modules. Mention your module mymod under appropriate header. This will discover all of the docstrings in mymod and create the appropriate webpage.

Building the Website

Building the website/documentation requires the following dependencies:

  1. Sphinx

  2. Cloud Sphinx Theme

  3. numpydoc

Note that xonsh itself needs to be installed too.

Procedure for modifying the website

The xonsh website source files are located in the docs directory. A developer first makes necessary changes, then rebuilds the website locally by executing the command:

$ make html

This will generate html files for the website in the _build/html/ folder. The developer may view the local changes by opening these files with their favorite browser, e.g.:

$ google-chrome _build/html/index.html

Once the developer is satisfied with the changes, the changes should be committed and pull-requested per usual. Once the pull request is accepted, the developer can push their local changes directly to the website by:

$ make push-root

Branches and Releases

Mainline xonsh development occurs on the main branch. Other branches may be used for feature development (topical branches) or to represent past and upcoming releases.

All releases should have a release candidate (‘-rc1’) that comes out 2 - 5 days prior to the scheduled release. During this time, no changes should occur to a special release branch (‘vX.X.X-release’).

The release branch is there so that development can continue on the develop branch while the release candidates (rc) are out and under review. This is because otherwise any new developments would have to wait until post-release to be merged into develop to prevent them from accidentally getting released early.

As such, the ‘vX.X.X-release’ branch should only exist while there are release candidates out. They are akin to a temporary second level of staging, and so everything that is in this branch should also be part of main.

Every time a new release candidate comes out the vX.X.X-release should be tagged with the name ‘X.X.X-rcX’. There should be a 2 - 5 day period of time in between release candidates. When the full and final release happens, the ‘vX.X.X-release’ branch is merged into main and then deleted.

If you have a new fix that needs to be in the next release candidate, you should make a topical branch and then pull request it into the release branch. After this has been accepted, the topical branch should be merged with main as well.

The release branch must be quiet and untouched for 2 - 5 days prior to the full release.

The release candidate procedure here only applies to major and minor releases. Micro releases may be pushed and released directly without having a release candidate.

Maintenance Tasks

You can cleanup your local repository of transient files such as *.pyc files created by unit testing by running:

$ rm -f xonsh/ xonsh/
$ rm -f xonsh/*.pyc tests/*.pyc
$ rm -fr build

Performing the Release

This is done through the rever. To get a list of the valid options use:

$ pip install re-ver

You can perform a full release:

$ rever check
$ rever <version-number>

Cross-platform testing

Most of the time, an actual VM machine is needed to test the nuances of cross platform testing. But alas here are some other ways to test things

  1. Windows

  • wine can be used to emulate the development environment. It provides cmd.exe with its default installation.

  1. OS X

  • darlinghq can be used to emulate the development environment for Linux users. Windows users can use Linux inside a virtual machine or WSL to run the same.

  • OSX KVM <> can be used for virtualization.

  1. Linux

  • It far easier to test things for Linux. docker is available on all three platforms.

One can leverage the Github Actions to provide a reverse shell to test things out. Solutions like actions-tmate are available, but they should not in any way violate the Github Action policies.

Document History

Portions of this page have been forked from the PyNE documentation, Copyright 2011-2015, the PyNE Development Team. All rights reserved.