• Stars
    star
    802
  • Rank 54,691 (Top 2 %)
  • Language
    Python
  • License
    MIT License
  • Created almost 9 years ago
  • Updated 3 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

the blessed package to manage your versions by scm tags

setuptools_scm

setuptools_scm extracts Python package versions from git or hg metadata instead of declaring them as the version argument or in a SCM managed file.

Additionally setuptools_scm provides setuptools with a list of files that are managed by the SCM (i.e. it automatically adds all of the SCM-managed files to the sdist). Unwanted files must be excluded by discarding them via MANIFEST.in.

setuptools_scm supports the following scm out of the box:

  • git
  • mercurial
https://tidelift.com/badges/package/pypi/setuptools-scm

pyproject.toml usage

The preferred way to configure setuptools_scm is to author settings in a tool.setuptools_scm section of pyproject.toml.

This feature requires Setuptools 42 or later, released in Nov, 2019. If your project needs to support build from sdist on older versions of Setuptools, you will need to also implement the setup.py usage for those legacy environments.

First, ensure that setuptools_scm is present during the project's built step by specifying it as one of the build requirements.

# pyproject.toml
[build-system]
requires = ["setuptools>=45", "setuptools_scm[toml]>=6.2"]

That will be sufficient to require setuptools_scm for projects that support PEP 518 (pip and pep517). Many tools, especially those that invoke setup.py for any reason, may continue to rely on setup_requires. For maximum compatibility with those uses, consider also including a setup_requires directive (described below in setup.py usage and setup.cfg).

To enable version inference, you need to set the version dynamically in the project section of pyproject.toml:

# pyproject.toml
[project]
# version = "0.0.1"  # Remove any existing version parameter.
dynamic = ["version"]

Then add this section to your pyproject.toml:

# pyproject.toml
[tool.setuptools_scm]

Including this section is comparable to supplying use_scm_version=True in setup.py. Additionally, include arbitrary keyword arguments in that section to be supplied to get_version(). For example:

# pyproject.toml
[tool.setuptools_scm]
write_to = "pkg/_version.py"

Where pkg is the name of your package.

If you need to confirm which version string is being generated or debug the configuration, you can install setuptools-scm directly in your working environment and run:

$ python -m setuptools_scm

# To explore other options, try:
$ python -m setuptools_scm --help

setup.py usage (deprecated)

Warning

setup_requires has been deprecated in favor of pyproject.toml

The following settings are considered legacy behavior and superseded by the pyproject.toml usage, but for maximal compatibility, projects may also supply the configuration in this older form.

To use setuptools_scm just modify your project's setup.py file like this:

  • Add setuptools_scm to the setup_requires parameter.
  • Add the use_scm_version parameter and set it to True.

For example:

from setuptools import setup
setup(
    ...,
    use_scm_version=True,
    setup_requires=['setuptools_scm'],
    ...,
)

Arguments to get_version() (see below) may be passed as a dictionary to use_scm_version. For example:

from setuptools import setup
setup(
    ...,
    use_scm_version = {
        "root": "..",
        "relative_to": __file__,
        "local_scheme": "node-and-timestamp"
    },
    setup_requires=['setuptools_scm'],
    ...,
)

You can confirm the version number locally via setup.py:

$ python setup.py --version

Note

If you see unusual version numbers for packages but python setup.py --version reports the expected version number, ensure [egg_info] is not defined in setup.cfg.

setup.cfg usage (deprecated)

as setup_requires is deprecated in favour of pyproject.toml usage in setup.cfg is considered deprecated, please use pyproject.toml whenever possible.

Programmatic usage

In order to use setuptools_scm from code that is one directory deeper than the project's root, you can use:

from setuptools_scm import get_version
version = get_version(root='..', relative_to=__file__)

See setup.py Usage (deprecated) above for how to use this within setup.py.

Retrieving package version at runtime

If you have opted not to hardcode the version number inside the package, you can retrieve it at runtime from PEP-0566 metadata using importlib.metadata from the standard library (added in Python 3.8) or the importlib_metadata backport:

from importlib.metadata import version, PackageNotFoundError

try:
    __version__ = version("package-name")
except PackageNotFoundError:
    # package is not installed
    pass

Alternatively, you can use pkg_resources which is included in setuptools (but has a significant runtime cost):

from pkg_resources import get_distribution, DistributionNotFound

try:
    __version__ = get_distribution("package-name").version
except DistributionNotFound:
     # package is not installed
    pass

However, this does place a runtime dependency on setuptools and can add up to a few 100ms overhead for the package import time.

Usage from Sphinx

It is discouraged to use setuptools_scm from Sphinx itself, instead use importlib.metadata after editable/real installation:

# contents of docs/conf.py
from importlib.metadata import version
release = version('myproject')
# for example take major/minor
version = '.'.join(release.split('.')[:2])

The underlying reason is, that services like Read the Docs sometimes change the working directory for good reasons and using the installed metadata prevents using needless volatile data there.

Usage from Docker

By default, docker will not copy the .git folder into your container. Therefore, builds with version inference might fail. Consequently, you can use the following snippet to infer the version from the host os without copying the entire .git folder to your Dockerfile.

RUN --mount=source=.git,target=.git,type=bind \
    pip install --no-cache-dir -e .

However, this build step introduces a dependency to the state of your local .git folder the build cache and triggers the long-running pip install process on every build. To optimize build caching, one can use an environment variable to pretend a pseudo version that is used to cache the results of the pip install process:

FROM python
COPY pyproject.toml
ARG PSEUDO_VERSION=1
RUN SETUPTOOLS_SCM_PRETEND_VERSION=${PSEUDO_VERSION} pip install -e .[test]
RUN --mount=source=.git,target=.git,type=bind pip install -e .

Note that running this Dockerfile requires docker with BuildKit enabled [docs].

To avoid BuildKit and mounting of the .git folder altogether, one can also pass the desired version as a build argument. Note that SETUPTOOLS_SCM_PRETEND_VERSION_FOR_${NORMALIZED_DIST_NAME} is preferred over SETUPTOOLS_SCM_PRETEND_VERSION.

Default versioning scheme

In the standard configuration setuptools_scm takes a look at three things:

  1. latest tag (with a version number)
  2. the distance to this tag (e.g. number of revisions since latest tag)
  3. workdir state (e.g. uncommitted changes since latest tag)

and uses roughly the following logic to render the version:

no distance and clean:
{tag}
distance and clean:
{next_version}.dev{distance}+{scm letter}{revision hash}
no distance and not clean:
{tag}+dYYYYMMDD
distance and not clean:
{next_version}.dev{distance}+{scm letter}{revision hash}.dYYYYMMDD

The next version is calculated by adding 1 to the last numeric component of the tag.

For Git projects, the version relies on git describe, so you will see an additional g prepended to the {revision hash}.

Semantic Versioning (SemVer)

Due to the default behavior it's necessary to always include a patch version (the 3 in 1.2.3), or else the automatic guessing will increment the wrong part of the SemVer (e.g. tag 2.0 results in 2.1.devX instead of 2.0.1.devX). So please make sure to tag accordingly.

Note

Future versions of setuptools_scm will switch to SemVer by default hiding the the old behavior as an configurable option.

Builtin mechanisms for obtaining version numbers

  1. the SCM itself (git/hg)
  2. .hg_archival files (mercurial archives)
  3. .git_archival.txt files (git archives, see subsection below)
  4. PKG-INFO

Git archives

Git archives are supported, but a few changes to your repository are required.

Create a .git_archival.txt file in the root directory of your repository, and copy-paste this into it:

node: $Format:%H$
node-date: $Format:%cI$
describe-name: $Format:%(describe:tags=true,match=*[0-9]*)$
ref-names: $Format:%D$

Create the .gitattributes file in the root directory of your repository if it doesn't already exist, and copy-paste this into it:

.git_archival.txt  export-subst

Finally, don't forget to commit those two files:

git add .git_archival.txt .gitattributes && git commit

Note that if you are creating a _version.py file, note that it should not be kept in version control.

File finders hook makes most of MANIFEST.in unnecessary

setuptools_scm implements a file_finders entry point which returns all files tracked by your SCM. This eliminates the need for a manually constructed MANIFEST.in in most cases where this would be required when not using setuptools_scm, namely:

  • To ensure all relevant files are packaged when running the sdist command.
  • When using include_package_data to include package data as part of the build or bdist_wheel.

MANIFEST.in may still be used: anything defined there overrides the hook. This is mostly useful to exclude files tracked in your SCM from packages, although in principle it can be used to explicitly include non-tracked files too.

Configuration parameters

In order to configure the way use_scm_version works you can provide a mapping with options instead of a boolean value.

The currently supported configuration keys are:

root:

Relative path to cwd, used for finding the SCM root; defaults to .

version_scheme:

Configures how the local version number is constructed; either an entrypoint name or a callable.

local_scheme:

Configures how the local component of the version is constructed; either an entrypoint name or a callable.

write_to:

A path to a file that gets replaced with a file containing the current version. It is ideal for creating a _version.py file within the package, typically used to avoid using pkg_resources.get_distribution (which adds some overhead).

Warning

Only files with .py and .txt extensions have builtin templates, for other file types it is necessary to provide write_to_template.

write_to_template:

A newstyle format string that is given the current version as the version keyword argument for formatting.

relative_to:

A file from which the root can be resolved. Typically called by a script or module that is not in the root of the repository to point setuptools_scm at the root of the repository by supplying __file__.

tag_regex:
A Python regex string to extract the version part from any SCM tag.

The regex needs to contain either a single match group, or a group named version, that captures the actual version information.

Defaults to the value of setuptools_scm.config.DEFAULT_TAG_REGEX (see _config.py).

parentdir_prefix_version:

If the normal methods for detecting the version (SCM version, sdist metadata) fail, and the parent directory name starts with parentdir_prefix_version, then this prefix is stripped and the rest of the parent directory name is matched with tag_regex to get a version string. If this parameter is unset (the default), then this fallback is not used.

This is intended to cover GitHub's "release tarballs", which extract into directories named projectname-tag/ (in which case parentdir_prefix_version can be set e.g. to projectname-).

fallback_version:

A version string that will be used if no other method for detecting the version worked (e.g., when using a tarball with no metadata). If this is unset (the default), setuptools_scm will error if it fails to detect the version.

parse:

A function that will be used instead of the discovered SCM for parsing the version. Use with caution, this is a function for advanced use, and you should be familiar with the setuptools_scm internals to use it.

git_describe_command:

This command will be used instead the default git describe command. Use with caution, this is a function for advanced use, and you should be familiar with the setuptools_scm internals to use it.

Defaults to the value set by setuptools_scm.git.DEFAULT_DESCRIBE (see git.py).

normalize:

A boolean flag indicating if the version string should be normalized. Defaults to True. Setting this to False is equivalent to setting version_cls to setuptools_scm.version.NonNormalizedVersion

version_cls:

An optional class used to parse, verify and possibly normalize the version string. Its constructor should receive a single string argument, and its str should return the normalized version string to use. This option can also receive a class qualified name as a string.

This defaults to packaging.version.Version if available. If packaging is not installed, pkg_resources.packaging.version.Version is used. Note that it is known to modify git release candidate schemes.

The setuptools_scm.NonNormalizedVersion convenience class is provided to disable the normalization step done by packaging.version.Version. If this is used while setuptools_scm is integrated in a setuptools packaging process, the non-normalized version number will appear in all files (see write_to) BUT note that setuptools will still normalize it to create the final distribution, so as to stay compliant with the python packaging standards.

To use setuptools_scm in other Python code you can use the get_version function:

from setuptools_scm import get_version
my_version = get_version()

It optionally accepts the keys of the use_scm_version parameter as keyword arguments.

Example configuration in setup.py format:

from setuptools import setup

setup(
    use_scm_version={
        'write_to': '_version.py',
        'write_to_template': '__version__ = "{version}"',
        'tag_regex': r'^(?P<prefix>v)?(?P<version>[^\+]+)(?P<suffix>.*)?$',
    }
)

Environment variables

SETUPTOOLS_SCM_PRETEND_VERSION:

when defined and not empty, its used as the primary source for the version number in which case it will be an unparsed string

SETUPTOOLS_SCM_PRETEND_VERSION_FOR_${NORMALIZED_DIST_NAME}:
 

when defined and not empty, its used as the primary source for the version number in which case it will be an unparsed string

the dist name normalization follows adapted PEP-503 semantics, with one or more of ".-_" being replaced by a single "_", and the name being upper-cased

it takes precedence over SETUPTOOLS_SCM_PRETEND_VERSION

SETUPTOOLS_SCM_DEBUG:

when defined and not empty, a lot of debug information will be printed as part of setuptools_scm operating

SOURCE_DATE_EPOCH:

when defined, used as the timestamp from which the node-and-date and node-and-timestamp local parts are derived, otherwise the current time is used (https://reproducible-builds.org/docs/source-date-epoch/)

SETUPTOOLS_SCM_IGNORE_VCS_ROOTS:

when defined, a os.pathsep separated list of directory names to ignore for root finding

Extending setuptools_scm

setuptools_scm ships with a few setuptools entrypoints based hooks to extend its default capabilities.

Adding a new SCM

setuptools_scm provides two entrypoints for adding new SCMs:

setuptools_scm.parse_scm
A function used to parse the metadata of the current workdir using the name of the control directory/file of your SCM as the entrypoint's name. E.g. for the built-in entrypoint for git the entrypoint is named .git and references setuptools_scm.git:parse

The return value MUST be a setuptools_scm.version.ScmVersion instance created by the function setuptools_scm.version:meta.

setuptools_scm.files_command

Either a string containing a shell command that prints all SCM managed files in its current working directory or a callable, that given a pathname will return that list.

Also use then name of your SCM control directory as name of the entrypoint.

Version number construction

setuptools_scm.version_scheme

Configures how the version number is constructed given a setuptools_scm.version.ScmVersion instance and should return a string representing the version.

Available implementations:

guess-next-dev:Automatically guesses the next development version (default). Guesses the upcoming release by incrementing the pre-release segment if present, otherwise by incrementing the micro segment. Then appends .devN. In case the tag ends with .dev0 the version is not bumped and custom .devN versions will trigger a error.
post-release:generates post release versions (adds .postN)
python-simplified-semver:Basic semantic versioning. Guesses the upcoming release by incrementing the minor segment and setting the micro segment to zero if the current branch contains the string 'feature', otherwise by incrementing the micro version. Then appends .devN. Not compatible with pre-releases.
release-branch-semver:Semantic versioning for projects with release branches. The same as guess-next-dev (incrementing the pre-release or micro segment) if on a release branch: a branch whose name (ignoring namespace) parses as a version that matches the most recent tag up to the minor segment. Otherwise if on a non-release branch, increments the minor segment and sets the micro segment to zero, then appends .devN.
no-guess-dev:Does no next version guessing, just adds .post1.devN
setuptools_scm.local_scheme

Configures how the local part of a version is rendered given a setuptools_scm.version.ScmVersion instance and should return a string representing the local version. Dates and times are in Coordinated Universal Time (UTC), because as part of the version, they should be location independent.

Available implementations:

node-and-date:adds the node on dev versions and the date on dirty workdir (default)
node-and-timestamp:like node-and-date but with a timestamp of the form {:%Y%m%d%H%M%S} instead
dirty-tag:adds +dirty if the current workdir has changes
no-local-version:omits local version, useful e.g. because pypi does not support it

Importing in setup.py

To support usage in setup.py passing a callable into use_scm_version is supported.

Within that callable, setuptools_scm is available for import. The callable must return the configuration.

# content of setup.py
import setuptools

def myversion():
    from setuptools_scm.version import get_local_dirty_tag
    def clean_scheme(version):
        return get_local_dirty_tag(version) if version.dirty else '+clean'

    return {'local_scheme': clean_scheme}

setup(
    ...,
    use_scm_version=myversion,
    ...
)

Customizing Version Scheme with pyproject.toml

To support custom version schemes in pyproject.toml, you may specify your own function as an entrypoint for getting the version.

# pyproject.toml
[tool.setuptools_scm]
version_scheme = "myproject.my_file:myversion_func"
# myproject/my_file
def myversion_func(version: ScmVersion):
    from setuptools_scm.version import guess_next_version
    return version.format_next_version(guess_next_version, '{guessed}b{distance}')

Note on testing non-installed versions

While the general advice is to test against a installed version, some environments require a test prior to install,

$ python setup.py egg_info
$ PYTHONPATH=$PWD:$PWD/src pytest

Interaction with Enterprise Distributions

Some enterprise distributions like RHEL7 and others ship rather old setuptools versions due to various release management details.

In those case its typically possible to build by using a sdist against setuptools_scm<2.0. As those old setuptools versions lack sensible types for versions, modern setuptools_scm is unable to support them sensibly.

In case the project you need to build can not be patched to either use old setuptools_scm, its still possible to install a more recent version of setuptools in order to handle the build and/or install the package by using wheels or eggs.

Code of Conduct

Everyone interacting in the setuptools_scm project's codebases, issue trackers, chat rooms, and mailing lists is expected to follow the PSF Code of Conduct.

Security Contact

To report a security vulnerability, please use the Tidelift security contact. Tidelift will coordinate the fix and disclosure.

More Repositories

1

pipenv

Python Development Workflow for Humans.
Python
24,500
star
2

pip

The Python package installer
Python
9,178
star
3

pipx

Install and Run Python Applications in Isolated Environments
Python
8,438
star
4

hatch

Modern, extensible Python project management
Python
5,078
star
5

sampleproject

A sample project that exists for PyPUG's "Tutorial on Packaging and Distributing Projects"
Python
4,914
star
6

virtualenv

Virtual Python Environment builder
Python
4,642
star
7

pipfile

Python
3,231
star
8

setuptools

Official project repository for the Setuptools build system
Python
2,251
star
9

flit

Simplified packaging of Python modules
Python
2,070
star
10

cibuildwheel

🎡 Build Python wheels for all the platforms with minimal configuration.
Python
1,684
star
11

twine

Utilities for interacting with PyPI
Python
1,543
star
12

manylinux

Python wheels that work on any linux (almost)
Shell
1,365
star
13

packaging.python.org

Python Packaging User Guide
Python
1,362
star
14

pip-audit

Audits Python environments and dependency trees for known vulnerabilities
Python
888
star
15

gh-action-pypi-publish

The blessed :octocat: GitHub Action, for publishing your 📦 distribution files to PyPI: https://github.com/marketplace/actions/pypi-publish
Python
795
star
16

get-pip

Helper scripts to install pip, in a Python installation that doesn't have it.
Python
694
star
17

build

A simple, correct Python build frontend
Python
667
star
18

packaging

Core utilities for Python packages
Python
566
star
19

wheel

The official binary distribution format for Python
Python
469
star
20

bandersnatch

A PyPI mirror client according to PEP 381 http://www.python.org/dev/peps/pep-0381/
Python
421
star
21

auditwheel

Auditing and relabeling cross-distribution Linux wheels.
Python
403
star
22

advisory-database

Advisory database for Python packages published on pypi.org
236
star
23

python-manylinux-demo

Demo project for building Python wheels for Linux with Travis-CI
C
221
star
24

sample-namespace-packages

Tests against namespace packages
Python
166
star
25

readme_renderer

Safely render long_description/README files in Warehouse
Python
151
star
26

packaging-problems

An issue tracker for the problems in packaging
140
star
27

trove-classifiers

Canonical source for classifiers on PyPI.
Python
126
star
28

pyproject-hooks

A low-level library for calling build-backends in `pyproject.toml`-based project
Python
115
star
29

installer

A low-level library for installing from a Python wheel distribution.
Python
107
star
30

scripttest

Utilities to help with testing command line scripts
Python
59
star
31

gh-action-pip-audit

A GitHub Action for pip-audit
Python
59
star
32

distlib

A low-level library which implements some Python packaging standards (PEPs) and which could be used by third-party packaging tools to achieve interoperability.
Python
43
star
33

distutils

distutils as found in cpython
Python
42
star
34

.github

Community health files for the Python Packaging Authority
30
star
35

interoperability-peps

Development repo for evolution of PyPA interoperability standards (released versions are published as PEPs on python.org)
Python
22
star
36

pypa.io

Source code for the pypa.io website
Python
22
star
37

integration-test

ensure core packaging tools work well with each other
17
star
38

pkg_resources

Abandoned extraction of pkg_resources. Official version found at /pypa/setuptools.
Python
15
star
39

get-virtualenv

14
star
40

history

history generates history/changelog files for a project
Python
7
star
41

easy_install

Python
7
star
42

wheel-builders

Companion repo for the [email protected] mailing list
7
star
43

browntruck

Python
6
star
44

pypa-docs-theme

Common base Sphinx theme for PyPA projects
Python
6
star
45

docker-python

5
star
46

copr

Package files for building PyPA packages in copr
4
star
47

pep470

Python
4
star
48

pip-test-package

Used in pip's test suite
Python
4
star
49

pypa-bot

Source code behind @pypa-bot (eventually)
3
star
50

pypi-camo

Dockerfile
3
star
51

bot-test

1
star