Project interface

Our projects should all include a package.json file at the top-level of the project, which defines the standard ways the project should be run.

The package.json file should define some scripts, and (usually) some dependencies and devDependencies.

scripts

These defined scripts can be run either with yarn run {script-name} or through our local development tooling.

Our standard script names:

  • start (required): The standard entrypoint for local development. It should run any required watchers for rebuilding files as they change, and run any local development servers etc…
    • E.g.: concurrently 'yarn run watch' 'yarn run serve'
  • build: Manually build any required assets for the project to run. This may simply run a number of other scripts per filetype, e.g.:
    • build-js: Build the JavaScript files
    • build-css: Build any SCSS files etc. into CSS
  • serve: Run the a local webserver. This will typically use the same ./entrypoint script as is used by the production Docker image (see below).
  • watch: Watch for changes in any relevant files. Again, this may be split up into sup-scripts by filetype, e.g.:
    • watch-js: Watch for changes in JavaScript files and rebuild as needed.
    • watch-css: Watch for changes in SCSS files and rebuild CSS as needed.
  • test: Run all tests. This may be split into sub-scripts by filetype, e.g.:
    • test-js: JavaScript tests.
    • test-python: Python tests.
    • lint-js: Check JavaScript format.
    • lint-python: Check Python format with black and flake8.
    • lint-css: Check CSS format.
  • format: Reformat all files with our standard formatters. This may be split into sub-scripts by filetype, e.g.:
    • format-python: Format Python with black --line-length 79
    • format-js: Format JavaScript with Prettier
    • format-css: Format SCSS files with Prettier

Basically every project should define at least start. All website projects should define serve. Any project may define any number of further scripts, as needed.

Publishing websites and entrypoint

We publish our website projects by building Docker images from them for deploying to our Kubernetes cluster. We build these images with e.g.:

docker build --build-arg BUILD_ID={build-id} .

To enable this, each website projects should contain a Dockerfile at the root of the project. This Dockerfile should use multiple stages to do any necessary preparatory steps - e.g. installing dependencies and building assets.

All images must be built from an official ubuntu image. Images should run a service on port 80, which will typically be started through a file called ./entrypoint, e.g.:

FROM ubuntu:bionic

# ... other RUN, ADD commands etc.

# Setup commands to run server
ENTRYPOINT ["./entrypoint"]
CMD ["0.0.0.0:80"]

package.json, the node_modules folder and any other unused files should usually not be included or used in our published images, and so should either be ignored through .dockerignore or explicitly removed in the Dockerfile.

Last updated 9 months ago.

Help improve this document in the forum.