Development Mode (a.k.a. “Editable Installs”)

When creating a Python project, developers usually want to implement and test changes iteratively, before cutting a release and preparing a distribution archive.

In normal circumstances this can be quite cumbersome and require the developers to manipulate the PYTHONPATH environment variable or to continuously re-build and re-install the project.

To facilitate iterative exploration and experimentation, setuptools allows users to instruct the Python interpreter and its import machinery to load the code under development directly from the project folder without having to copy the files to a different location in the disk. This means that changes in the Python source code can immediately take place without requiring a new installation.

You can enter this “development mode” by performing an editable installation inside of a virtual environment, using pip’s -e/--editable flag, as shown below:

$ cd your-python-project
$ python -m venv .venv
# Activate your environemt with:
#      `source .venv/bin/activate` on Unix/macOS
# or   `.venv\Scripts\activate` on Windows

$ pip install --editable .

# Now you have access to your package
# as if it was installed in .venv
$ python -c "import your_python_project"

An “editable installation” works very similarly to a regular install with pip install ., except that it only installs your package dependencies, metadata and wrappers for console and GUI scripts. Under the hood, setuptools will try to create a special .pth file in the target directory (usually site-packages) that extends the PYTHONPATH or install a custom import hook.

When you’re done with a given development task, you can simply uninstall your package (as you would normally do with pip uninstall <package name>).

Please note that, by default an editable install will expose at least all the files that would be available in a regular installation. However, depending on the file and directory organization in your project, it might also expose as a side effect files that would not be normally available. This is allowed so you can iteratively create new Python modules. Please have a look on the following section if you are looking for a different behaviour.

Virtual Environments

You can think about virtual environments as “isolated Python runtime deployments” that allow users to install different sets of libraries and tools without messing with the global behaviour of the system.

They are a safe way of testing new projects and can be created easily with the venv module from the standard library.

Please note however that depending on your operating system or distribution, venv might not come installed by default with Python. For those cases, you might need to use the OS package manager to install it. For example, in Debian/Ubuntu-based systems you can obtain it via:

sudo apt install python3-venv

Alternatively, you can also try installing virtualenv. More information is available on the Python Packaging User Guide on Install packages in a virtual environment using pip and venv.

Note

Changed in version v64.0.0: Editable installation hooks implemented according to PEP 660. Support for namespace packages is still EXPERIMENTAL.

“Strict” editable installs

When thinking about editable installations, users might have the following expectations:

1. It should allow developers to add new files (or split/rename existing ones) and have them automatically exposed.

2. It should behave as close as possible to a regular installation and help users to detect problems (e.g. new files not being included in the distribution).

Unfortunately these expectations are in conflict with each other. To solve this problem setuptools allows developers to choose a more “strict” mode for the editable installation. This can be done by passing a special configuration setting via pip, as indicated below:

pip install -e . --config-settings editable_mode=strict

In this mode, new files won’t be exposed and the editable installs will try to mimic as much as possible the behavior of a regular install. Under the hood, setuptools will create a tree of file links in an auxiliary directory ($your_project_dir/build) and add it to PYTHONPATH via a .pth file. (Please be careful to not delete this repository by mistake otherwise your files may stop being accessible).

Warning

Strict editable installs require auxiliary files to be placed in a build/__editable__.* directory (relative to your project root).

Please be careful to not remove this directory while testing your project, otherwise your editable installation may be compromised.

You can remove the build/__editable__.* directory after uninstalling.

Note

New in version v64.0.0: Added new strict mode for editable installations. The exact details of how this mode is implemented may vary.

Limitations

• The editable term is used to refer only to Python modules inside the package directories. Non-Python files, external (data) files, executable script files, binary extensions, headers and metadata may be exposed as a snapshot of the version they were at the moment of the installation.

• Adding new dependencies, entry-points or changing your project’s metadata require a fresh “editable” re-installation.

• Console scripts and GUI scripts MUST be specified via entry-points to work properly.

• Strict editable installs require the file system to support either symbolic or hard links. This installation mode might also generate auxiliary files under the project directory.

• There is no guarantee that the editable installation will be performed using a specific technique. Depending on each project, setuptools may select a different approach to ensure the package is importable at runtime.

• There is no guarantee that files outside the top-level package directory will be accessible after an editable install.

• There is no guarantee that attributes like __path__ or __file__ will correspond to the exact location of the original files (e.g., setuptools might employ file links to perform the editable installation). Users are encouraged to use tools like importlib.resources or importlib.metadata when trying to access package files directly.

• Editable installations may not work with namespaces created with pkgutil or pkg_resources. Please use PEP 420-style implicit namespaces [1].

• Support for PEP 420-style implicit namespace packages for projects structured using flat-layout is still experimental. If you experience problems, you can try converting your package structure to the src-layout.

• File system entries in the current working directory whose names coincidentally match installed packages may take precedence in Python’s import system. Users are encouraged to avoid such scenarios [2].

Attention

Editable installs are not a perfect replacement for regular installs in a test environment. When in doubt, please test your projects as installed via a regular wheel. There are tools in the Python ecosystem, like tox or nox, that can help you with that (when used with appropriate configuration).

Legacy Behavior

If your project is not compatible with the new “editable installs” or you wish to replicate the legacy behavior, for the time being you can also perform the installation in the compat mode:

pip install -e . --config-settings editable_mode=compat

This installation mode will try to emulate how python setup.py develop works (still within the context of PEP 660).

Warning

The compat mode is transitional and will be removed in future versions of setuptools, it exists only to help during the migration period. Also note that support for this mode is limited: it is safe to assume that the compat mode is offered “as is”, and improvements are unlikely to be implemented. Users are encouraged to try out the new editable installation techniques and make the necessary adaptations.

If the compat mode does not work for you, you can also disable the editable install hooks in setuptools by setting an environment variable:

SETUPTOOLS_ENABLE_FEATURES="legacy-editable"

This may cause the installer (e.g. pip) to effectively run the “legacy” installation command: python setup.py develop [3].

How editable installations work

Advanced topic

There are many techniques that can be used to expose packages under development in such a way that they are available as if they were installed. Depending on the project file structure and the selected mode, setuptools will choose one of these approaches for the editable installation [4].

A non-exhaustive list of implementation mechanisms is presented below. More information is available on the text of PEP 660.

• A static .pth file [5] can be added to one of the directories listed in site.getsitepackages() or site.getusersitepackages() to extend sys.path.

• A directory containing a farm of file links that mimic the project structure and point to the original files can be employed. This directory can then be added to sys.path using a static .pth file.

• A dynamic .pth file [6] can also be used to install an “import finder” (MetaPathFinder or PathEntryFinder) that will hook into Python’s import system machinery.

Attention

Setuptools offers no guarantee of which technique will be used to perform an editable installation. This will vary from project to project and may change depending on the specific version of setuptools being used.

---

Notes

[1]

You may be able to use strict editable installations with namespace packages created with pkgutil or pkg_namespaces, however this is not officially supported.

[2]

Techniques like the src-layout or tooling-specific options like tox’s changedir can be used to prevent such kinds of situations (checkout this blog post for more insights).

[3]

For this workaround to work, the installer tool needs to support legacy editable installations. (Future versions of pip, for example, may drop support for this feature).

[4]

setuptools strives to find a balance between allowing the user to see the effects of project files being edited while still trying to keep the editable installation as similar as possible to a regular installation.

[5]

i.e., a .pth file where each line correspond to a path that should be added to sys.path. See Site-specific configuration hook.

[6]

i.e., a .pth file that starts where each line starts with an import statement and executes arbitrary Python code. See Site-specific configuration hook.