Skip to content

mtkennerly/poetry-dynamic-versioning

Repository files navigation

Dynamic versioning plugin for Poetry

This is a Python 3.7+ plugin for Poetry 1.2.0+ and Poetry Core 1.0.0+ to enable dynamic versioning based on tags in your version control system, powered by Dunamai. Many different version control systems are supported, including Git and Mercurial; please refer to the Dunamai page for the full list (and minimum supported version where applicable).

poetry-dynamic-versioning provides a build backend that patches Poetry Core to enable the versioning system in PEP 517 build frontends. When installed with the plugin feature (i.e., poetry-dynamic-versioning[plugin]), it also integrates with the Poetry CLI to trigger the versioning in commands like poetry build.

For Poetry 1.1.x, you can use an older version of poetry-dynamic-versioning (0.17.1 or earlier) that relied on a *.pth import hack, but this is no longer supported, so you should migrate to the standardized plugin and Poetry 1.2.0+.

Installation

If you've previously installed the deprecated poetry-dynamic-versioning-plugin package, be sure to uninstall it before proceeding.

  • Install the plugin using one of the options below:

    • In most cases: poetry self add "poetry-dynamic-versioning[plugin]"
    • If you installed Poetry with Pipx: pipx inject poetry "poetry-dynamic-versioning[plugin]"

    See the Poetry plugin documentation for more information about these options.

  • Run in your project: poetry dynamic-versioning enable

    Or you can update your pyproject.toml manually:

    [tool.poetry-dynamic-versioning]
    enable = true

    Include the plugin in the build-system section of pyproject.toml for interoperability with PEP 517 build frontends:

    [build-system]
    requires = ["poetry-core>=1.0.0", "poetry-dynamic-versioning>=1.0.0,<2.0.0"]
    build-backend = "poetry_dynamic_versioning.backend"

    This is a thin wrapper around poetry.core.masonry.api.

Poetry still requires the tool.poetry.version field to be present in pyproject.toml, but you are encouraged to use version = "0.0.0" as a standard placeholder.

With the minimal configuration above, the plugin will automatically take effect when you run commands such as poetry build. It will update the version in pyproject.toml, then revert the change when the plugin deactivates.

The default configuration will also update any pre-existing __version__ = "0.0.0" and __version_tuple__ = (0, 0, 0) placeholders in some files. You can configure additional substitution patterns/files as needed (see below).

Configuration

In your pyproject.toml file, you may configure the following options:

  • [tool.poetry-dynamic-versioning]: General options.

    • enable (boolean, default: false): Since the plugin has to be installed globally, this setting is an opt-in per project.

    • vcs (string, default: any): This is the version control system to check for a version. One of: any, git, mercurial, darcs, bazaar, subversion, fossil, pijul.

    • metadata (boolean, default: unset): If true, include the commit hash in the version, and also include a dirty flag if dirty is true. If unset, metadata will only be included if you are on a commit without a version tag. This is ignored when format or format-jinja is used.

    • tagged-metadata (boolean, default: false): If true, include any tagged metadata discovered as the first part of the metadata segment. Has no effect when metadata is set to false. This is ignored when format or format-jinja is used.

    • dirty (boolean, default: false): If true, include a dirty flag in the metadata, indicating whether there are any uncommitted changes. Has no effect when metadata is set to false. This is ignored when format or format-jinja is used.

    • pattern (string): This is a regular expression which will be used to find a tag representing a version. When this is unset, Dunamai's default pattern is used.

      There must be a capture group named base with the main part of the version. Optionally, it may contain another two groups named stage and revision for prereleases, and it may contain a group named tagged_metadata to be used with the tagged-metadata option. There may also be a group named epoch for the PEP 440 concept.

      If the base group is not included, then this will be interpreted as a named preset from the Dunamai Pattern class. This includes: default, default-unprefixed (makes the v prefix optional).

      You can check the default for your installed version of Dunamai by running this command:

      poetry run python -c "import dunamai; print(dunamai.Pattern.Default.regex())"
      

      Remember that backslashes must be escaped in the TOML file.

      # Regular expression:
      pattern = '(?P<base>\d+\.\d+\.\d+)'
      # Named preset:
      pattern = "default-unprefixed"
    • pattern-prefix (string): This will be inserted after the pattern's start anchor (^). For example, to match tags like some-package-v1.2.3, you can keep the default pattern and set the prefix to some-package-.

    • format (string, default: unset): This defines a custom output format for the version. Available substitutions:

      • {base}
      • {stage}
      • {revision}
      • {distance}
      • {commit}
      • {dirty}
      • {tagged_metadata}
      • {branch}
      • {branch_escaped} which omits any non-letter/number characters
      • {timestamp} of the current commit, which expands to YYYYmmddHHMMSS as UTC

      Example: v{base}+{distance}.{commit}

    • format-jinja (string, default: unset): This defines a custom output format for the version, using a Jinja template. When this is set, format is ignored.

      Available variables:

      • base (string)
      • stage (string or None)
      • revision (integer or None)
      • distance (integer)
      • commit (string)
      • dirty (boolean)
      • tagged_metadata (string or None)
      • version (dunumai.Version)
      • env (dictionary of environment variables)
      • branch (string or None)
      • branch_escaped (string or None)
      • timestamp (string or None)

      Available functions:

      Simple example:

      format-jinja = "{% if distance == 0 %}{{ base }}{% else %}{{ base }}+{{ distance }}.{{ commit }}{% endif %}"

      Complex example:

      format-jinja = """
          {%- if distance == 0 -%}
              {{ serialize_pep440(base, stage, revision) }}
          {%- elif revision is not none -%}
              {{ serialize_pep440(base, stage, revision + 1, dev=distance, metadata=[commit]) }}
          {%- else -%}
              {{ serialize_pep440(bump_version(base), stage, revision, dev=distance, metadata=[commit]) }}
          {%- endif -%}
      """
    • format-jinja-imports (array of tables, default: empty): This defines additional things to import and make available to the format-jinja template. Each table must contain a module key and may also contain an item key. Consider this example:

      format-jinja-imports = [
          { module = "foo" },
          { module = "bar", item = "baz" },
      ]

      This is roughly equivalent to:

      import foo
      from bar import baz

      foo and baz would then become available in the Jinja formatting.

    • style (string, default: unset): One of: pep440, semver, pvp. These are preconfigured output formats. If you set both a style and a format, then the format will be validated against the style's rules. If style is unset, the default output format will follow PEP 440, but a custom format will only be validated if style is set explicitly.

    • latest-tag (boolean, default: false): If true, then only check the latest tag for a version, rather than looking through all the tags until a suitable one is found to match the pattern.

    • bump (boolean, default: false): If true, then increment the last part of the version base by 1, unless the stage is set, in which case increment the revision by 1 or set it to a default of 2 if there was no revision. Does nothing when on a commit with a version tag.

      Example, if there have been 3 commits since the v1.3.1 tag:

      • PEP 440 with bump = false: 1.3.1.post3.dev0+28c1684
      • PEP 440 with bump = true: 1.3.2.dev3+28c1684
    • tag-dir (string, default: tags): This is the location of tags relative to the root. This is only used for Subversion.

    • tag-branch (string, default: unset): Branch on which to find tags, if different than the current branch. This is only used for Git currently.

    • full-commit (boolean, default: false): If true, get the full commit hash instead of the short form. This is only used for Git and Mercurial.

    • strict (boolean, default: false): If true, then fail instead of falling back to 0.0.0 when there are no tags.

    • fix-shallow-repository (boolean, default: false): If true, then automatically try to fix shallow repositories. Currently, this only supports Git and will run git fetch --unshallow.

    • ignore-untracked (boolean, default: false): If true, ignore untracked files when determining whether the repository is dirty.

  • [tool.poetry-dynamic-versioning.substitution]: Insert the dynamic version into additional files other than just pyproject.toml. These changes will be reverted when the plugin deactivates.

    • files (array of strings): Globs for any files that need substitutions. Default: ["*.py", "*/__init__.py", "*/__version__.py", "*/_version.py"]. To disable substitution, set this to an empty list.

    • patterns (array of strings/tables): Regular expressions for the text to replace. Each regular expression must have two capture groups, which are any text to preserve before and after the replaced text.

      String items are interpreted as a regular expression directly. Table items support these keys:

      • value (string): This is the regular expression.
      • mode (string, optional): This controls how the version should be inserted. Options:
        • str (default): Serialize version as-is. The capture groups should already include the surrounding quotation marks.
        • tuple: Serialize 0.1.2.dev0+a.b as 0, 1, 2, "dev0", "a.b". The capture groups should already include the surrounding parentheses.

      Default:

      patterns = [
          '''(^__version__\s*(?::.*?)?=\s*['"])[^'"]*(['"])''',
          { value = '''(^__version_tuple__\s*(?::.*?)?=\s*\()[^)]*(\))''', mode = "tuple" },
      ]

      Remember that the backslashes must be escaped in the TOML file.

    • folders (array of tables, default: empty): List of additional folders to check for substitutions.

      Each table supports these options:

      • path (string, required): Path to the folder.
      • files (array of strings, optional): Override substitution.files for this folder.
      • patterns (array of strings, optional): Override substitution.patterns for this folder.

      If you use an src layout, you may want to keep the default files/patterns and just specify the following:

      folders = [
        { path = "src" }
      ]

      This will check the default file globs (e.g., ./*.py) as well as the same file globs inside of src (e.g., ./src/*.py).

  • [tool.poetry-dynamic-versioning.files] (table, default: empty): This section lets you tweak the behavior for individual files. Each table key is a path to a specific file (no globs) relative to the project root. Each nested table supports these fields:

    If you use these options to generate a file that's ignored by your VCS, but you also want the generated file to be included in the output of poetry build, then you'll need to name the file explicitly in your tool.poetry.include config.

    • persistent-substitution (boolean, optional): If true, then do not revert any substitutions applied to this file. This is primarily useful for editable installs, if you need the version to remain in a file ignored by your VCS.

    • initial-content (string, optional): Set the file content before the substitution phase. The file will be created or overwritten as necessary. Common leading whitespace will be stripped from each line.

    • initial-content-jinja (string, optional): Same as initial-content, but using Jinja formatting. If both options are set, this one takes priority. You can use the same imports from format-jinja-imports and the same variables from format-jinja, with this additional variable:

      • formatted_version (string) - version formatted by either the format or format-jinja option

    Example:

    [tool.poetry-dynamic-versioning.files."package/_version.py"]
    persistent-substitution = true
    initial-content = """
      # These version placeholders will be replaced later during substitution.
      __version__ = "0.0.0"
      __version_tuple__ = (0, 0, 0)
    """
  • [tool.poetry-dynamic-versioning.from-file]: This section lets you read the version from a file instead of the VCS.

    • source (string): If set, read the version from this file. It must be a path relative to the location of pyproject.toml. By default, the plugin will read the entire content of the file, without leading and trailing whitespace.
    • pattern (string): If set, use this regular expression to extract the version from the file. The first capture group must contain the version.

Simple example:

[tool.poetry-dynamic-versioning]
enable = true
vcs = "git"
style = "semver"

Environment variables

In addition to the project-specific configuration above, you can apply some global overrides via environment variables.

  • POETRY_DYNAMIC_VERSIONING_BYPASS: Use this to bypass the VCS mechanisms and use a static version instead. The value of the environment variable will be used as the version for the active project and any path/SSH dependencies that also use the plugin. This is mainly for distro package maintainers who need to patch existing releases, without needing access to the original repository.
  • POETRY_DYNAMIC_VERSIONING_OVERRIDE: Use a static version for specific packages only, but leave dynamic versioning enabled otherwise. For example, pkg1 = 0.1.0, pkg2 = 0.2.0 (spaces are optional) would set pkg1 to 0.1.0 and pkg2 to 0.2.0. This only affects packages for which poetry-dynamic-versioning is enabled. When both variables are set, OVERRIDE takes precedence over BYPASS.
  • POETRY_DYNAMIC_VERSIONING_COMMANDS: You can set a comma-separated list of Poetry commands during which to activate the versioning. For example, build,publish will limit the dynamic versioning to those two commands. Similarly, setting POETRY_DYNAMIC_VERSIONING_COMMANDS="" will disable the dynamic versioning altogether, which is useful in isolated environments like Docker where versions may not be computable and not needed anymore.
  • POETRY_DYNAMIC_VERSIONING_COMMANDS_NO_IO: Comma-separated list of Poetry commands during which the plugin should not directly modify files. The plugin will still set the dynamic version in memory so that Poetry itself can write it as needed. Default: version.
  • POETRY_DYNAMIC_VERSIONING_DEBUG: If this is set to 1, then some debug logs will be printed to stderr. Right now, this logs some cases where substitution doesn't find anything to change.

Command line mode

The plugin also has a command line mode for execution on demand. This mode applies the dynamic version to all relevant files and leaves the changes in-place, allowing you to inspect the result. Your configuration will be detected from pyproject.toml as normal, but the enable option is not necessary.

To activate this mode, either use poetry dynamic-versioning (provided by the plugin feature) or poetry-dynamic-versioning (standalone script with default features).

VCS archives

Sometimes, you may only have access to an archive of a repository (e.g., a zip file) without the full history. The plugin can still detect a version in some of these cases. Refer to the Dunamai documentation for more info.

Caveats

All of Dunamai's caveats apply. In addition to those:

  • The dynamic version is not available during poetry run or poetry shell because of a Poetry design choice that prevents the plugin from cleaning up after itself.

  • Regarding PEP 517 support:

    pip wheel . and pip install . will work with new enough Pip versions that default to in-tree builds or support the --use-feature=in-tree-build option. Older versions of Pip will not work because they create an isolated copy of the source code that does not contain the version control history.

    If you want to build wheels of your dependencies, you can do the following, although local path-based dependencies may not work:

    poetry export -f requirements.txt -o requirements.txt --without-hashes
    pip wheel -r requirements.txt