Explicit, strict and automatic project version management based on semantic versioning.
MIT License
Explicit, strict and automatic project version management based on semantic versioning.
If you found this project useful, but it does not really fit your development and releasing processes, create an issue with your proposals, please.
Also, if you have any questions after reading the documentation, check FAQ. If there are no answers, create an issue with your question, please.
An end user of the project is a software or DevOps-related engineer who develop, release and deploy projects that need explicit, strict and automatic project version management for tags, images, API and/or libraries.
There is the semantic versioning. To make a long story short, it is versioning specification
with major, minor and patch numbers telling us what the current version of a project is and how to react on its new
versions' updates. Example of the project version following semantic versioning is 1.3.12
where the first digit is
major number, the second digit is minor number, and the third digit is patch number.
These are the rules of increasing chose numbers:
Increase major version when you make incompatible API changes such as change a name of a required function's or API parameter.
Before changes: 1.3.12
After changes: 2.0.0
Increase minor version when you add functionality in a backwards compatible manner such as add a new optional function's or API parameter.
Before changes: 1.3.12
After changes: 1.4.0
Increase match version when you make backwards compatible bug fixes.
Before changes: 1.3.12
After changes: 1.3.13
You have probably seen semantic versioning in your programming language's packages such as
JavaScript's axios (e.g. version 0.24.0
) or
Python requests (e.g. version 2.27.1
).
project version
is just a set of principles to maintain project versioning and
command line interface that helps not to forget about those principles such as a code style
and linters to check its compliance.
project version
requires having a file named .project-version
in the root directory containing a project version.
With this file, developers declare single source to fetch a project version from for things like Git
tags or Docker
images.
Now you can reuse a project version from .project-version
file for multiple release-related purposes:
There may be situations when you deployed the new version of an application but do not have deployment logs, or you
deployed the new version of a application but logs tell nothing, or do not have Git
information. In all the cases
it may be useful to enter your application's runtime and check file .project-version
to know the exact version
which is related to the codebase.
$ cat .project-version
1.3.12
Instead of using Git
commit SHA or its short version for Docker
images, you can use a project version.
$ docker build --tag facebook/react:v$(cat .project-version) -f Dockerfile .
Instead of using Git
commit SHA or its short version for GitHub
release version number, you can use a project
version.
on:
push:
branches:
- master
jobs:
release:
runs-on: [ubuntu-latest]
steps:
- uses: actions/checkout@v2
- name: Create Release
uses: actions/create-release@v1
with:
tag_name: $(cat .project-version)
release_name: Release v$(cat .project-version)
Instead of supporting package version (Python package
, Gem
or JAR
) in a dedicated file, you can automatically
use a project version. Python package
with its setup.py
for building packages is illustrated below:
with open('.project-version', 'r') as project_version_file:
project_version = project_version_file.read().strip()
setup(
version=project_version,
name='project-version',
...
)
In case you manage an infrastructure as a code (e.g. Kubernetes
), you may face challenges of supporting multiple
major version of your project (e.g. HTTP API
). Without automation, you should create new major version
configurations manually.
Let us consider the example where you have API's first version deployment configurations:
$ ls deployment/
├── deployment
│ ├── v1
│ │ └── deployment.yaml
│ │ └── ingress.yaml
│ │ └── service.yaml
$ cat /deployment/v1/deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
namespace: api-v1
When it is time to create API's second version, you can simply copy previous version configurations and substitute
v1
to v2
.
$ echo .project-version
2.0.0
$ export PROJECT_PREVIOUS_MAJOR_VERSION=$(($cat .project-version)-1)
$ echo $PROJECT_MAJOR_VERSION
1
$ export PROJECT_MAJOR_VERSION=$(cut -d '.' -f 1 <<< "$(cat .project-version)")
$ echo $PROJECT_PREVIOUS_MAJOR_VERSION
2
$ cp -r \
deployment/v$PROJECT_PREVIOUS_MAJOR_VERSION deployment/v$PROJECT_MAJOR_VERSION
$ find deployment/ -type f -exec sed -i \
's/namespace: v$PROJECT_PREVIOUS_MAJOR_VERSION/namespace: v$PROJECT_MAJOR_VERSION/g' {} +
After, you automatically get deployment configurations for the new version:
$ ls deployment/
├── deployment
│ ├── v1
│ │ └── deployment.yaml
│ │ └── ingress.yaml
│ │ └── service.yaml
│ ├── v2
│ │ └── deployment.yaml
│ │ └── ingress.yaml
│ │ └── service.yaml
$ cat /deployment/v2/deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
namespace: api-v2
And there is much more cases when relying on a project version from its file makes software releasing easier.
All use cases described above requires a project version always be up-to-date and never corrupted. In case it is not,
you can release the same version twice, for example. To avoid this, project-version
is tightly bound to a branching
model with its release life-cycle. Let's consider how project-version
works with the most popular branching models
Git flow
and GitHub flow
.
In Git flow
, developers do features in feature branches and merge them to develop
branch. When develop
branch
has a set of features merged, a release is created (with a separate branch for it) and deployed. To define a release
version, project-version
requests a developer to make an additional commit into develop
branch that changes
.project-version
file.
In GitHub flow
, developers do features in feature branches and merge them to develop
branch. Once a single feature
is merged to develop
branch, a release is immediately created (with no separate branch) and deployed. To define a
release version, project-version
requests a developer to make an additional commit into a feature branch that changes
.project-version
file.
This chapter describes a set of command line interface (automation scripts) with descriptive explanation of its use-cases that help to manage a project version. The command line interface is completely optional but helpful. It helps developers not to forget about increasing a project version or auto-increase when needed.
Install using pip3
:
$ pip3 install project-version
Get the version of the package — project-version --version
:
$ project-version --version
project-version, version 0.1.0
Get the detailed description of all supported commands by the package — project-version --help
:
$ project-version --help
Usage: project-version [OPTIONS] COMMAND [ARGS]...
Project version command-line interface.
Options:
--version Show the version and exit.
--help Show this message and exit.
Commands:
check Check whether specified project version is increased properly.
Check whether specified project version is increased properly — project-version check
.
Environment variables:
Variable | Type | Required | Restrictions | Description |
---|---|---|---|---|
ACCESS_TOKEN | String | Yes | - | The provider's API access token. |
Parameters:
Argument | Type | Required | Restrictions | Description |
---|---|---|---|---|
provider | String | Yes | One of: GitHub . |
A provider of hosting for software development and version control name. |
organization | String | Yes | - | The provider's organization name. |
repository | String | Yes | - | The provider's repository name. |
base-branch | String | Yes | - | A branch to compare a project version with. Usually, a default branch. |
head-branch | String | Yes | - | A branch to get its project version for comparison. Usually, a feature branch. |
Example of usage:
$ project-version check \
--provider=GitHub \
--organization=facebook \
--repository=react \
--base-branch=master \
--head-branch=map-children-components
A use case:
The use case of the command is to prevent merging feature branches to develop
or master
that have not increased
a project version. As illustrated below: if both master
and feature branch have project version 1.1.3
, the command
exits with failed status code, if feature branch has 1.2.0
and master
has 1.1.3
(lower version), the command exits
with succeed status code.
The example of a failed pipeline:
The example of a pipeline configuration:
---
name: Pull request workflow
on:
pull_request_target:
branches:
- master
jobs:
check-project-version:
runs-on: [ubuntu-latest]
steps:
- uses: actions/checkout@v2
- name: Install project version
run: pip3 install project-version
- name: Check a project version
env:
ACCESS_TOKEN: ${{secrets.GIT_HUB_ACCESS_TOKEN}}
run: |
project-version check \
--provider=GitHub \
--organization=facebook \
--repository=react \
--base-branch=master \
--head-branch=map-children-components
Bump the minor version of a project version — project-version bump
.
Environment variables:
Variable | Type | Required | Restrictions | Description |
---|---|---|---|---|
ACCESS_TOKEN | String | Yes | - | The provider's API access token. |
Parameters:
Argument | Type | Required | Restrictions | Description |
---|---|---|---|---|
provider | String | Yes | One of: GitHub . |
A provider of hosting for software development and version control name. |
organization | String | Yes | - | The provider's organization name. |
repository | String | Yes | - | The provider's repository name. |
base-branch | String | Yes | - | A branch to get a project version from. Usually, a default branch. |
head-branch | String | Yes | - | A branch to push bumped project version to. Usually, a feature branch. |
The example of usage:
$ project-version bump \
--provider=GitHub \
--organization=facebook \
--repository=react \
--base-branch=master \
--head-branch=dependabot/npm/core-js-3.6.4
A use case:
There are tools like Dependabot that automatically do updates to your codebase.
Dependabot
tracks your requirements' versions and keep them up-to-date through proposing pull requests. The example
of changes is illustrated below:
Usually, codebase is covered with tests and there is no need to review such small changes and developers setup these
pull requests to be merged automatically. But it will be impossible with project version
as it requires a project
version to be increased manually. The bump command can be applied here, you just configure that if a pull request
is opened by Dependabot
and then execute the bumping command.
The example of a pipeline configuration:
---
name: Pull request workflow
on:
pull_request_target:
branches:
- master
jobs:
check-project-version:
runs-on: [ubuntu-latest]
steps:
- uses: actions/checkout@v2
- name: Install project version
run: pip3 install project-version
- name: Bump project version if it is non-human pull request
if: ${{ github.actor == 'dependabot[bot]' || github.actor == 'facebook-bot' }}
env:
ACCESS_TOKEN: ${{secrets.GIT_HUB_ACCESS_TOKEN}}
run: |
project-version bump \
--provider=GitHub \
--organization=facebook \
--repository=react \
--base-branch=master \
--head-branch=map-children-components
Make a release — project-version release
.
Environment variables:
Variable | Type | Required | Restrictions | Description |
---|---|---|---|---|
ACCESS_TOKEN | String | Yes | - | The provider's API access token. |
Parameters:
Argument | Type | Required | Restrictions | Description |
---|---|---|---|---|
provider | String | Yes | One of: GitHub . |
A provider of hosting for software development and version control name. |
organization | String | Yes | - | The provider's organization name. |
repository | String | Yes | - | The provider's repository name. |
branch | String | Yes | - | A branch to make a release for |
project-version | String | Yes | - | A project version to make a release with. |
access-token | String | Yes | - | The provider's API access token. |
The example of usage:
$ project-version release \
--provider=GitHub \
--organization=dmytrostriletskyi \
--repository=project-version \
--branch=master \
--project-version=1.1.3
A use case:
When you squash merge
your feature branches, your develop
or master
commits history might looks like:
Allow taxi drivers to schedule a break (#567)
Integrate Spotify (#566)
Create OpenAPI spification (#565)
If you pick one and create a release with a commit as the title, you will have Allow taxi drivers to schedule a break (#567)
.
If you add a release version, it will go uglier like Release v3.6.1: Allow taxi drivers to schedule a break (#567)
.
The release command remove all the mess and create the following title v3.6.1: allow taxi drivers to schedule a break
.
The example of a pipeline configuration:
---
name: Master workflow
on:
push:
branches:
- master
jobs:
release:
runs-on: [ubuntu-latest]
outputs:
project_version: ${{ steps.get_project_version.outputs.project_version }}
steps:
- uses: actions/checkout@v2
- name: Install project version
run: pip3 install project-version
- name: Get a version of the project
id: get_project_version
run: echo "::set-output name=project_version::$(cat .project-version)"
- name: Release
env:
ACCESS_TOKEN: ${{secrets.GIT_HUB_ACCESS_TOKEN}}
run: |
project-version release \
--provider=GitHub \
--organization=facebook \
--repository=react \
--branch=master \
--project-version=${{ steps.get_project_version.outputs.project_version }}
Q: project-version
is written in Python
, but my project's stack is different. Why should I support Python` for
this?
A: When you develop a project, you do not need Python
, but only .project-version
file. The only place you need
Python
on is your pipelines runner such as GitHub Actions
, Jenkins
or GitLab CI/CD
to run the command line
interface. You can use isolated environment such as Docker
containers:
jobs:
check-project-version:
runs-on: [ubuntu-latest]
container:
image: python:3.9.0-slim
...
Q: Why should a developer increase a project version manually for a feature, or a set of features?
A: When a developer does a change, the only they know a degree of change: either patch, minor or major. There is no machine learning model or other software that can describe a degree of change instead of a person who made those changes.
Q: If we merge feature branches often, many concurrent feature branches should pull new project version often. Is it fine?
A: Yes, it is fine. It is a price you pay for the project management. Also, keep in mind that most time you develop a feature, and only little time you pull other feature branches' changes and merge.
Clone the project and install requirements:
$ git clone [email protected]:dmytrostriletskyi/accessify.git && cd accessify
$ make install-requirements
After changes, ensure the code quality remains the same:
$ make check-requirements-safety
$ make check-code-complexity
$ make check-code-quality
$ make check-yaml-standards
If you are new for the contribution, please read: