Home

Awesome

Build

See the configuration guide for help getting started, selecting a versioning strategy and example configurations, or contributing.md for information on how to get help or contribute to this project.

Git-Based Semantic Versioning

This action produces a semantic version for a repository using the repository's git history without ever requiring a human to choose or manually assign the version number.

This action is designed to facilitate assigning version numbers during a build automatically while publishing version that only increment by one value per release. To accomplish this, the next version number is calculated along with a commit increment indicating the number of commits for this version. The commit messages are inspected to determine the type of version change the next version represents. Including the term (MAJOR) or (MINOR) in the commit message alters the type of change the next version will represent.

Background

Automatic versioning during a build presents a chicken-and-egg problem--we want the version to increase by a single value between each release, but we usually do not know at build time whether a new build will be released or not. Generally a build is tagged as part of a release step after passing testing and other quality controls, so if we want to use the version number in the build itself, especially for a build triggered by a commit, we cannot rely on having a proper tag for the build. Most CI systems offer a "build number", but this does not correspond to our semantic version and relies on the state of the CI tool. It is with this in mind that this tool was developed with the following goals:

To solve this problem, this action calculates the next implied version based on the most recently tagged version and the commit messages. An additional value called the "increment" tracks the count of commits since the last version change, allowing a label to be created to mark pre-release versions. The version produced by this action is always the implied version (unless bump_each_commit is set to true). Subsequently tagging a commit that is chosen as the implied version is what bumps the version for future commits.

Commits Graph

Unless the current commit is already tagged, the version produced by this action will be one value ahead of the last tag.

Major and Minor Versions

The commit messages for the span of commits from the last tag are checked for the presence of the designated terms ((MAJOR) or (MINOR) by default), if a term is encountered that commit is treated as the start of a major or minor version instead of the default patch level. As with normal commits the implied version will only increment by one value since the last tag regardless of how many major or minor commits are encountered. Major commits override minor commits, so a set of commits containing both major and minor tags will result in a major version increment.

Commits Graph

Tags on Previous Commits

Adding a tag to an older commit changes the implicit version of commits since the tagged commit. If a tag is assigned to an older commit, the commits that come after it will be given the new version if the build were to be retriggered, for example:

Commits Graph

Usage

<!-- start usage -->
- uses: paulhatch/semantic-version@v5.4.0
  with:
    # The prefix to use to identify tags
    tag_prefix: "v"
    # A string which, if present in a git commit, indicates that a change represents a
    # major (breaking) change, supports regular expressions wrapped with '/'
    major_pattern: "(MAJOR)"
    # A string which indicates the flags used by the `major_pattern` regular expression. Supported flags: idgs
    major_regexp_flags: ""
    # Same as above except indicating a minor change, supports regular expressions wrapped with '/'
    minor_pattern: "(MINOR)"
    # A string which indicates the flags used by the `minor_pattern` regular expression. Supported flags: idgs
    minor_regexp_flags: ""
    # A string to determine the format of the version output
    version_format: "${major}.${minor}.${patch}-prerelease${increment}"
    # Optional path to check for changes. If any changes are detected in the path the
    # 'changed' output will true. Enter multiple paths separated by spaces.
    change_path: "src/my-service"
    # Named version, will be used as suffix for name version tag
    namespace: my-service
    # If this is set to true, *every* commit will be treated as a new version.
    bump_each_commit: false
    # If bump_each_commit is also set to true, setting this value will cause the version to increment only if the pattern specified is matched.
    bump_each_commit_patch_pattern: ""
    # If true, the body of commits will also be searched for major/minor patterns to determine the version type.
    search_commit_body: false
    # The output method used to generate list of users, 'csv' or 'json'.
    user_format_type: "csv"
    # Prevents pre-v1.0.0 version from automatically incrementing the major version.
    # If enabled, when the major version is 0, major releases will be treated as minor and minor as patch. Note that the version_type output is unchanged.
    enable_prerelease_mode: true
    # If enabled, diagnostic information will be added to the action output.
    debug: false
    # If true, the branch will be used to select the maximum version.
    version_from_branch: false

Outputs

There are two types of "version" string, one is the semantic version output that can be used to identify a build and can include prerelease data and metadata specific to the commit such as v2.0.1-pre001+cf6e75 (you would produce this string yourself using the version information from this action plus whatever metadata you wanted to add), the other is the tag version string, which identifies a specific commit as being a specific version.

Using Multiple Versions in the Same Repository

It is possible to create additional versions for multiple project co-existing in one repository, for example you may have a Helm chart, database migration, or simply be hosting multiple projects in the same repository and want them to be versioned independently. There are a few settings that can be used to accomplish this:

First, you can set the change_path input to specify a path that will be inspected for changes. Commits which do no change any files in this path will not increase the increment output. In addition, if there are no changes in a given commit with this path specified, the changed value will be false.

Second, the input namespace can be set to create an additional named version. If this value is set, it will be appended (separated by a hyphen) to the end of tags for the version, and only tags with this value appended will be considered when determining the version. The namespace will be pruned from the string output as "version" within the action.

Finally, set different values for major_pattern and minor_pattern than the other projects in order to be able to mark these commits independently.

To use secondary versions in a workflow, simply create additional steps in a job referencing semantic version multiple times. For example, a project tagged like v1.2.3+0-db could be configured like this:

- name: Application Version
  id: version
  uses: paulhatch/semantic-version@v5.4.0
  with:
    change_path: "src/service"
- name: Database Version
  id: db-version
  uses: paulhatch/semantic-version@v5.4.0
  with:
    major_pattern: "(MAJOR-DB)"
    minor_pattern: "(MINOR-DB)"
    change_path: "src/migrations"
    namespace: db

Important Note Regarding the Checkout action

Beginning in v2, actions/checkout does not include tags/history by default. This history is required to determine the version correctly. To include the history and tags, specify the fetch-depth parameter in your checkout action declaration. Specify zero to pull the full history and tags.

  - name: Checkout
    uses: actions/checkout@v2
    with:
      fetch-depth: 0

Alternatively, you can set this number to a value high enough to pull all the commits you'd expect to have in a release.