This project does stuff with medical statistics, and serves as my primary portfolio project.

• Getting Started
  • Prerequisites
  • Setup
• Repo Components
• Application Services
• Developer Experience
  • Github Actions
  • Static Typing / MyPy
  • Dependabot

The following needs to be installed:

• docker compose (Any version should work. Tested with v1 and v2 standalone. Should also work with docker compose CLI plugin, but this is not tested.)
• bash. Trivial on most systems, but installing is important for Windows users.
• python >= 3.5. Needed for running management utilities.
• pip. Required to install a few packages used by the management utilities.

I have taken great care to ensure this project is as easy to get started with as humanly possible.

• Ensure prerequisites are installed.
• Install management packages via pip3 install -r requirements.txt
• python3 manager.py build to build docker images for the frontend and backend.
• python3 manager.py up to start the services. python3 manager.py down can be used later to shut them down.
• python3 manager.py init_db to (re)build the database, including running all migrations. If it already exists, it will be dropped, recreated, and migrated.

See README in commands folder for a complete list of commands and their descriptions

This section contains a description of all the major folders in the root of the repo.

The following tools and conventions are used to provide the best possible developer experience, and maximize system maintainability

GitHub Actions is used for the following purposes on this project:

• Performing CI checks to enforce various code quality metrics and conditions (eg, tests passing, linting, static analysis)
• Ensuring builds are easily repeatable and correctly automated
• Continuous updating of production as main gets updated
• Automatically and continously deploying successful builds of open pull requests to https://pr{{PR NUMBER HERE}}.medistat.online
  • Link to the deployed branch-under-review is automatically posted as a comment by a bot, when all checks are successful
  • Deployments for all open PRs get continuously updated until closed
  • Once PRs are merged or closed, their resources are all freed.

Precommit hooks give a certain level of quality-at-the-source by checking for easy to detect code issues before allowing commits. In most cases, and where possible, the pre-commit hooks will also correct the issues they find. Precommit hooks are used on this project for:

• Auto-formatting python, yaml, and json files. Checks against excessively large files being commited, trailing whitespace, lines at end of file, tabs, and CLRF.
• Checks that commited python code has a valid AST (ie, no python syntax errors)
• Enforcing static typing according to the mypy configuration
  • Since this was a legacy project that I revived for the sake of demonstrating instrastructure skills, the mypy configuration reflects that of a project that is seeing incremental adoption of static typing. New project areas have stricter rules enforced than older project areas.
• Checks to prevent breakpoint() from being commited

When a pull request is opened, or the branch in question receives new commits, the following occurs automatically via GitHub Actions:

• Code is installed on the application server at ~/medistat-pr{{PR_NUMBER_HERE}}.
• .env file is generated that contains all necessary config to start up this PR as a new docker stack without interfering with any other services
• Build process runs and produces new custom images if successful. If not successful, the developer who opened the PR is notified.
• Docker stack is started. Checks run to ensure that all stack services become healthy within a certain configurable timeframe.
• Unit and integration tests run. Additional code-quality checks are run.
• Database is built from scratch and data fixtures are loaded in.
• GitHub Actions bot posts the link to the deployed stack's frontend service, and any other admin services. (TODO: Generate credentials for every service and post the credentials on the PR). The URLs will follow the pattern of: https://pr{{PR_NUMBER_HERE}}.medistat.online and https://{{SERVICE_NAME_HERE}}pr{{PR_NUMBER_HERE}}.medistat.online

When a pull request is closed (usually because it was merged), the resources used are cleaned up:

• The docker stack is brought down
• The containers used by the PR's stack are deleted
• The directory containing all project files is deleted

When the main branch receives a push, GitHub Actions activates the continuous deployment script. This follows a very similar process to the Pull Request Workflow described above, with a few differences.

• Existing .env file is used; deploys to https://medistat.online instead of a subdomain of medistat.online
• Existing database is used, instead of creating a new database from scratch. New database migrations run.
• Production deployment happens under a different OS user than any staging/review deployment.