This middleware is used by Knative Functions to expose a Function written in Python as a network service.
.
├── cmd
│ └── fhttp - Example Function using the http middleware
│ └── fce - Example Function using the CloudEvent middleware
├── src/func_python
│ ├── http.py - HTTP Middleware
│ ├── cloudevents.py - CloudEvent Middleware
├── tests
│ ├── http_test.py - HTTP Middleware tests
│ ├── cloudevent_test.py - CloudEvent tests
└── README.md - This Readme
Run suite:
poetry run pytest
To enable more granular log levels:
poetry run pytest --log-cli-level=INFO
Minimal example of running the test command, which shows how this library is used when integrate with Functions, and can be useful during dev.
- install
poetry
viapipx
- install dependencies with
poetry install
- activate the virtual environment managed by poetry via
poetry shell
Note that in some environments this command may cause collissions with configured keyboard shortcuts. If there are problems, you can instead source the environment variables directly with:source $(poetry env info --path)/bin/activate
- run the example via
poetry run python cmd/fhttp/main.py
- deactivate the virtual environment with
exit
A nice method of development using git worktrees:
p. From a personal fork, create a new worktree for the bug, feature or chore
named appropriately (eg. "feature-a")
git worktree add feature-a
-
Implement the code changes and commit.
-
Update the CHANGELOG.md to include the change in the "Unreleased" section.
-
Commit, push and create a PR to the upstream repository's main branch.
-
Solicit a code-read from another team member.
-
Upon approval, squash and merge to main.
-
(optional) cleanup by removing the worktree and associated local and remote branches.
-
(optional) pull into local fork's main and push to remote fork main.
To run the package level tests
To test if the package published to Test PyPI works, ensure the project has it added as an explicit source in pyproject.toml:
[[tool.poetry.source]]
name = "test-pypi"
url = "https://test.pypi.org/simple/"
priority = "explicit"
Then update the dependency to explicitly pull the new version which is only available on TestPyPI. For example to test a hypothetical unreleased version 0.1.2:
[tool.poetry.dependencies]
func_python = {version = "0.1.2", source = "test-pypi"}
Run poetry install
to install the unreleased version from Test PyPI
Note: do not check in this change.
NOTE: This process is currently undergoing minor tweaks, and thus contains several manual steps. Once the process is codified, it can be more highly automated.
-
Create a new branch "release-x.y" from an up-to-date main When incrementing, follow Semantic Versioning standard. eg:
- Bug fixes: ++ patch version
- Features: ++ minor version
- Breaking Changes: ++ major version
-
Update pyproject.toml with the new version.
-
Update CHANGELOG.md by moving the "Unreleased" items into a new section for the given version (leaving an empty Unreleased section as a template for future updates)
-
Commit, push and create a PR to upstream's main branch Please set the commit message to "Release vx.y.z" with the new version number
-
Obtain approval for release from another team member.
-
Squash and merge (ensure the commit message remains Release vx.y.x).
-
Verify the new version was correctly published to test PyPI and all precautions have been taken to ensure it is functioning as intended. (See testing above)
-
Pull from upstream into your local main branch and tag the commit vx.y.z (the squash and merge will have created a new commit hash)
-
push the tag to upstream, triggering the release to production PyPI.
-
Update the GitHub release's notes to be the changelog section's contents .
Update to using the suffix "rcX" so as not to confuse folks pulling tags list from the git repo. Without this suffix, a simple listing of the latest tags from the repo would show a potentially unreleased version as the latest.
With the use of these release-candidate tags, the use of TestPyPI may be unnecessary.
Perhaps push the RC tag to the PR's commit and test prior to merging the PR to main, such that a potentially broken version is not kept in the main branch's history. For example, if v0.1.2rc1 failed, its commit will not be part of the main branch.