Ideally the description in the Jira task is written in a way that it can be copied and pasted as the description of the Pull Request. This is very subjective, but the description should:

• Clearly describe the change that is needed

• Be written in plain language - if it’s very technical then perhaps add a “Background” introduction that explains things in plain language

• Contain an “Acceptance Criteria” section detailing what is needed for the task to be considered done

There will always be improvements underway in the CHEFS code, and any new work must align with those improvements. The end goal of these improvements is code that is:

• necessary (YAGNI - don’t maintain, document, test, etc, code that we don’t need)

• correct (it should work!)

• simple (“as simple as possible but no simpler”)

• documented

• easy to read

• easy to understand

• easy to modify

• easy to fully test

• consistent (to reduce cognitive load)

Information that everyone should be familiar with:

• Contributing Guidelines

• CHEFS Technical Documentation

• OpenAPI Documentation for the backend

• Backend Design Document

• Backend Test Strategy

The goal for all developers is to create Pull Requests that are easy and fast to review. If a task contains a large amount of work, break it down into smaller stand-alone tasks. Jira subtasks are one way of recording this, and allow the subtask to go through the various Jira swimlanes.

• build: change in build system or dependencies

• ci: change in continuous integration / deployment

• docs: change to documentation

• feat: a new feature is being added

• fix: an existing bug or defect is being fixed

• perf: change to improve performance

• refactor: change to improve code quality

• revert: reverts changes in a previous commit

• style: change to code style/formatting

• test: add missing tests or correct existing tests

The fetch and rebase GitHub workflow used by Kubernetes is a common way to keep cloned repos in sync with the origin and upstream repos.

Note: Some people prefer to pull and create merge commits, and that’s fine because we squash commits when a Pull Request is merged into upstream/main.

You can name the branch anything you want, but:

• the feat/ prefix groups branches of the same type in the tools

• the task number 1234 makes it easy to find the corresponding ticket in Jira

It’s only a recommendation that “working” commit messages use this format - these commit messages will eventually be squashed, and it’s only the final commit message that must be in this format. It’s a good idea to always use this format to be familiar with how the final commit message must look.

Tip: In the VS Code commit message input box, the up arrow allows you to scroll back through previous commit messages

Test coverage reports appear in:

• Frontend: app/frontend/coverage/index.html

• Backend: app/coverage/lcov-report/index.html

Refer to the backend unit test documentation for details on:

• test writing strategy

• running tests on the command line

• running with and without the coverage report

• running single specs or single tests

We use the format type: FORMS-NNNN description for our Pull Request titles so that:

• the type of change is obvious from the name of the Pull Request

• it is easy to find all Pull Requests for a given type like feat or test

• it is easy to find the Pull Request for a given FORMS-NNNN Jira task

• the description makes it easy to scan Pull Requests to find a change that happened

This is a chance to do a self-review of your changes before creating the PR.

• are the changes too big (too many files, too many lines) to be easily and quickly reviewed?

• is the Pull Request doing too many different things? Can it be broken down into smaller PRs?

• is every changed line necessary? Was anything accidentally committed?

• are there tests to cover the code changes / additions?

• are there documentation updates for the code changes / additions?

The “Tests” GitHub Action runs:

• GitHub CodeQL tests to look for security issues

• Backend Jest unit tests, including test coverage saved as an artifact

• Frontend Jest unit tests, including test coverage saved as an artifact

Note that the test coverage is not uploaded to Code Climate, as Actions run from Pull Requests do not have access to the GitHub Secrets needed to authenticate with Code Climate.

There are various ways to tell this:

• Watching the Action you will see the Deploy step complete and display the URL of the deployment

• Unless you have the notifications disabled, you should receive an email

• The Action will add a comment to the PR that contains the URL of the deployment

Pull Requests should be created so that can be easily and quickly reviewed. By reading through every line of your code changes (because the reviewer should also be reading every line) you check that the Pull Request will make sense to your reviewers. It might help to add a GitHub comment to your code changes to explain them better. It might help to split your Pull Request into smaller pieces if it is too large: small Pull Requests that take five or ten minutes to review will get a more thorough review than Pull Requests that take hours or days.

Ideally every developer would review and understand every change in every Pull Request, but that’s not practical. Code reviews have many purposes:

• Sharing knowledge: more people understanding the changes means better group knowledge of the code

• Enforcing consistency: the code should look like a team effort, not a dozen people doing things their own way - this reduces cognitive load when working with the code

• Preventing bugs: the goal of a review is not to test that someone’s code works, that’s the developer’s job. Reviewers very familiar with the code being changed, though, are likely to catch bugs introduced by people less familiar with the code - this is a great learning opportunity.

There are two ways of looking at Pull Requests:

1. Those that use existing concepts / designs vs. those that introduce new concepts / designs

2. Those that are simple changes vs. those that are complex changes

We can use the Ship / Show / Ask approach to Pull Requests:

• Existing concepts / designs + simple changes: Ship it (example: dependency update, typo fix, small bug). Show or Ask for a review if you are uncertain or otherwise feel it’s necessary

• Existing concepts / designs + complex changes: Show / tell other developers about the changes but doesn’t necessarily need a review (example: new feature that’s very similar to existing features). Ask for a review if you are uncertain or otherwise feel it’s necessary

• New concepts / designs + simple changes: Ask for a review (example: small bugfix that needs a big new way of doing things)

• New concepts / designs + complex changes: Ask for a review (example: new feature that’s different from anything we have)

This step really depends on the reviewer. Some people notice that they have been asked to do a review, but many do not. It doesn’t hurt to post in the team channel on Discord that something is ready for review and “mention” the reviewers.

It depends! It’s good to always be up to date because the testing and review is more meaningful if the base branch is up to date. However, it can also waste a lot of time if one person is putting in multiple small changes and everyone is trying to keep up. Find a balance between having a current base and spending time keeping it current.

If you or others are testing the deployed code, but the code is not up to date with main, then bugs could sneak through. When you look at your PR there will be comments in the timeline for commits, and comments from the PR Deploy Action. If there are commits after the last PR Deploy, then it should be deployed again. Retesting might be wanted if there were merge conflicts or other contention between the new main and the PR code.

This format must be followed as we eventually want to introduce automated changelogs and versioning.

The Push Action runs on merge, and:

1. Build the application

2. Deploys the application to dev

3. Waits for gatekeeper approval to deploy to test

4. Deploys the application to test

5. Waits for gatekeeper approval to deploy to prod

6. Deploys the application to prod

Once you have merged your changes, ensure that they are deployed through all environments. Do not leave a deployment “hanging” and deployed to only the lower environments, as the next merge will probably cause it to go all the way to prod. If the change is tested and approved, it must go to prod.

Maybe? Merge can only be done by people with write access to the repo, so in theory on merge we would have access to the GitHub secrets and could undeploy automatically if we can figure out the PR number.

We use Code Climate to monitor maintainability and code coverage. However, it will probably be replaced by SonarCloud in the near future. But it doesn’t hurt to check that your changes didn’t decrease our maintainability or code coverage.