mirrors
/
pypa-hatch
mirror of https://github.com/pypa/hatch.git
0
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
pypa-hatch/docs/plugins/environment/virtual.md

6.9 KiB
Raw Permalink Blame History

Virtual environment

This uses virtual environments backed by virtualenv or UV.

Configuration

The environment plugin name is virtual.

[tool.hatch.envs.<ENV_NAME>]
type = "virtual"

Options

Option Default Description
python The version of Python to find on your system and subsequently use to create the environment, defaulting to the HATCH_PYTHON environment variable, followed by the normal resolution logic. Setting the HATCH_PYTHON environment variable to self will force the use of the Python executable Hatch is running on. For more information, see the documentation.
python-sources ['external', 'internal'] This may be set to an array of strings that are either the literal internal or external. External considers only Python executables that are already on PATH. Internal considers only internally managed Python distributions.
path An explicit path to the virtual environment. The path may be absolute or relative to the project root. Any environments that inherit this option will also use this path. The environment variable HATCH_ENV_TYPE_VIRTUAL_PATH may be used, which will take precedence.
system-packages false Whether or not to give the virtual environment access to the system site-packages directory
installer pip When set to uv, UV will be used in place of virtualenv & pip for virtual environment creation and dependency management, respectively. If you intend to provide UV yourself, you may set the HATCH_ENV_TYPE_VIRTUAL_UV_PATH environment variable which should be the absolute path to a UV binary. This environment variable implicitly sets the installer option to uv (if unset).

Location

The location of environments is determined in the following heuristic order:

  1. The path option
  2. A directory named after the environment within the configured virtual environment directory if the directory resides somewhere within the project root or if it is set to a .virtualenvs directory within the user's home directory
  3. Otherwise, environments are stored within the configured virtual environment directory in a deeply nested structure in order to support multiple projects

Additionally, when the path option is not used, the name of the directory for the default environment will be the normalized project name to provide a more meaningful default shell prompt.

Python resolution

Virtual environments necessarily require a parent installation of Python. The following rules determine how the parent is resolved.

The Python choice is determined by the python option followed by the HATCH_PYTHON environment variable. If the choice is via the environment variable, then resolution stops and that path is used unconditionally.

The resolvers will be based on the python-sources option and all resolved interpreters will ensure compatibility with the project's defined Python support.

If a Python version has been chosen then each resolver will try to find an interpreter that satisfies that version.

If no version has been chosen, then each resolver will try to find a version that matches the version of Python that Hatch is currently running on. If not found then each resolver will try to find the highest compatible version.

!!! note Some external Python paths are considered unstable and are ignored during resolution. For example, if Hatch is installed via Homebrew then sys.executable will be ignored because the interpreter could change or be removed at any time.

!!! note When resolution finds a match using an internally managed distribution and an update is available, the latest distribution will automatically be downloaded before environment creation.

Internal distributions

The following options are recognized for internal Python resolution.

!!! tip You can set custom sources for distributions by setting the HATCH_PYTHON_SOURCE_<NAME> environment variable where <NAME> is the uppercased version of the distribution name with periods replaced by underscores e.g. HATCH_PYTHON_SOURCE_PYPY3_10.

CPython

NAME
3.7
3.8
3.9
3.10
3.11
3.12
3.13

The source of distributions is the python-build-standalone project.

Some distributions have variants that may be configured with environment variables. Options may be combined.

Option Platforms Allowed values
HATCH_PYTHON_VARIANT_CPU
  • Linux
  • v1
  • v2
  • v3 (default)
  • v4
HATCH_PYTHON_VARIANT_GIL
  • Linux
  • Windows
  • macOS
  • freethreaded

PyPy

NAME
pypy2.7
pypy3.9
pypy3.10

The source of distributions is the PyPy project.

Troubleshooting

macOS SIP

If you need to set linker environment variables like those starting with DYLD_ or LD_, any executable secured by System Integrity Protection that is invoked when running commands will not see those environment variable modifications as macOS strips those.

Hatch interprets such commands as shell commands but deliberately ignores such paths to protected shells. This workaround suffices for the majority of use cases but there are 2 situations in which it may not:

  1. There are no unprotected sh, bash, zsh, nor fish executables found along PATH.
  2. The desired executable is a project's script, and the location of environments contains spaces or is longer than 1241 characters. In this case pip and other installers will create such an entry point with a shebang pointing to /bin/sh (which is protected) to avoid shebang limitations. Rather than changing the location, you could invoke the script as e.g. python -m pytest (if the project supports that method of invocation by shipping a __main__.py).

  1. The shebang length limit is usually 127 but 3 characters surround the executable path: #!<EXE_PATH>\n ↩︎