Skip to content

Development Guide

Welcome! Thank you for wanting to make the project better. This section provides an overview of repository structure and how to work with the code base.

Before you dive into this, it is best to read:

Docker

The Dagger project uses Docker to ease setting up a consistent development environment. The Docker documentation has details on how to install docker on your computer.

Once that is configured, the integration test suite can be run locally:

docker-compose run --rm integation_test 

Testing

You'll be unable to merge code unless the linting and tests pass. You can run these in your container via:

docker-compose run --rm test

This will run the same tests, linting, and code coverage that are run by the CI pipeline. The only difference is that, when run locally, black and isort are configured to automatically correct issues they detect.

Generally we should endeavor to write tests for every feature. Every new feature branch should increase the test coverage rather than decreasing it.

We use pytest as our testing framework.

Stages

To customize / override a specific testing stage, please read the documentation specific to that tool:

  1. PyTest
  2. MyPy
  3. Black
  4. Isort
  5. Flake8
  6. Bandit

Building the Library

dagger is PEP 517 compliant. build is used as the frontend tool for building the library. Setuptools is used as the build backend. setup.cfg contains the library metadata. A setup.py is also included to support an editable install.

Requirements

  • requirements.txt - Lists all direct dependencies (packages imported by the library).
  • requirements-test.txt - Lists all direct dependencies needed for development. This primarily covers dependencies needed to run the test suite & lints.

Publishing a New Version

Once the package is ready to be released, there are a few things that need to be done:

  1. Start with a local clone of the repo on the default branch with a clean working tree.
  2. Run the version bump script with the appropriate part name (major, minor, or patch). Example: docker-compose run --rm bump minor This wil create a new branch, updates all affected files with the new version, and commit the changes to the branch.

  3. Push the new branch to create a new pull request.

  4. Get the pull request approved.
  5. Merge the pull request to the default branch.

Merging the pull request will trigger a GitHub Action that will create a new release. The creation of this new release will trigger a GitHub Action that will to build a wheel & a source distributions of the package and push them to PyPI.

Warning

The action that uploads the files to PyPI will not run until a repository maintainer acknowledges that the job is ready to run. This is to keep the PyPI publishing token secure. Otherwise, any job would have access to the token.

In addition to uploading the files to PyPI, the documentation website will be updated to include the new version. If the new version is a full release, it will be made the new latest version.

Continuous Integration Pipeline

The Continuous Integration (CI) Pipeline runs to confirm that the repository is in a good state. It will run when someone creates a pull request or when they push new commits to the branch for an existing pull request. The pipeline runs multiple different jobs that helps verify the state of the code.

This same pipeline also runs on the default branch when a maintainer merges a pull request.

Lints

The first set of jobs that run as part of the CI pipline are linters that perform static analysis on the code. This includes: MyPy, Black, Isort, Flake8, and Bandit.

Tests

The next set of jobs run the unit tests using PyTest. The pipeline runs the tests cases across each supported version of Python to ensure compatibility.

For each run of the test cases, the job will record the test results and code coverage information. The pipeline uploads the code coverage information to CodeCov to ensure that a pull request doesn't significantly reduce the total code coverage percentage or introduce a large amount of code that is untested.

Distribution Verification

The next set of jobs build the wheel distribution, installs in into a virtual environment, and then runs Python to import the library version. This works as a smoke test to ensure that the library can be packaged correctly and used. The pipeline runs the tests cases across each supported version of Python to ensure compatibility.

Documentation

The remaining jobs are all related to documentation.

  • A job builds the documentation in strict mode so that it will fail if there are any errors. The job records the generated files so that the documentation website can be viewed in its rendered form.
  • When the pipeline is running as a result of a maintainer merging a pull request to the default branch, a job runs that publishes the current state of the documentation to as the dev version. This will allow users to view the state of the documentation as it has changed since a maintainer published the latest version.