• Repository

Meta

• License Expression: MIT
  SPDX License Expression
• Author: David Glick
• Requires: Python >=3.8

Classifiers

• Development Status
  • 4 - Beta
• Programming Language
  • Python
  • Python :: 3.8
  • Python :: 3.9
  • Python :: 3.10
  • Python :: 3.11
  • Python :: 3.12
  • Python :: 3.13
  • Python :: Implementation :: CPython
  • Python :: Implementation :: PyPy

🐎 Horse with no namespace

👁 PyPI - Version
👁 PyPI - Python Version

horse-with-no-namespace is a tool which converts pkg_resources namespace packages to be compatible with PEP 420 implicit namespace packages.

Why is it needed?

The Python ecosystem provides several implementations of namespace packages, which make it possible to split the modules within a single package to be distributed in multiple distribution packages.

See https://packaging.python.org/en/latest/guides/packaging-namespace-packages/ for more background on namespace packages and their various implementations.

The pkg_resources-style namespace packages are considered obsolete, but it is difficult to incrementally migrate a namespace using them to native namespace packages, because there are situations where not all of the packages will be found. For example, this happens if one package within the namespace uses pkg_resources-style namespace packages but another package uses one of the other styles, and they are not installed in the same path.

The goal of horse-with-no-namespace is to provide a temporary solution that can be installed to make things work during the interim period while existing releases of some packages in a namespace still use the pkg_resources style, but other packages are already in the process of being converted to native namespaces packages. Once all packages in the namespace have been converted, horse-with-no-namespace is no longer needed.

How does it work?

horse-with-no-namespace interferes with both of the ways that pkg_resources-style namespace packages can be loaded.

• For packages that are installed into a site-packages directory, there is a file ending in -nspkg.pth which is automatically loaded by Python's site module to create the namespace package in sys.modules. horse-with-no-namespace removes these namespace packages from sys.modules, so that the namespace packages can be reloaded later without using the .pth machinery. (horse-with-no-namespace actually uses a similar hack with its own .pth file to do this, but arranges for it to be loaded later.)
• For packages imported from somewhere else in sys.path, pkg_resources-style namespace packages are set up by a call to pkg_resources.declare_namespace from __init__.py. horse-with-no-namespace replaces pkg_resources.declare_namespace with a shim that instead initializes a pkgutil-style namespace package (which is compatible with native namespace packages).

Can I use it with Buildout?

This approach can also help with packages installed using zc.buildout. However, horse-with-no-namespace itself must not be installed using Buildout, but should be installed using pip into the same virtualenv where zc.buildout is installed. If horse-with-no-namespace is installed using Buildout, then its .pth file will not be loaded, since it is not in the site-packages folder.

How can I suppress the log line that is emitted to stderr on startup?

Set the environment variable PYTHONWARNINGS=ignore:::horse_with_no_namespace.

Installation

pip install horse-with-no-namespace

License

horse-with-no-namespace is distributed under the terms of the MIT license.

Verified details

Project links

• Documentation
• Issues
• Source

Owner

• Plone

GitHub Statistics

Maintainers

Unverified details

• Download URL: horse_with_no_namespace-20260202.0.tar.gz
• Upload date: Feb 3, 2026
• Size: 6.1 kB
• Tags: Source
• Uploaded using Trusted Publishing? Yes
• Uploaded via: twine/6.1.0 CPython/3.12.9

See more details on using hashes here.

• Download URL: horse_with_no_namespace-20260202.0-py3-none-any.whl
• Upload date: Feb 3, 2026
• Size: 6.2 kB
• Tags: Python 3
• Uploaded using Trusted Publishing? Yes
• Uploaded via: twine/6.1.0 CPython/3.12.9

See more details on using hashes here.