spreadsheet-api-python

This is intended to serve as an API backend with pluggable storage for a web-based spreadsheet application. I've chosen to build a spreadsheet application since it is easy to start simply, and later layer in complexity (e.g. data formatting, calculated values, websockets, concurrent editing).

This application will be implemented as a set of individual repositories and components in order to allow the use of various languages and frameworks. This also mimics a microservices system architecture, and will allow experimentation with different testing and deployment strategies.

While this is the first, I hope to include the following components:

• spreadsheet-frontend-svelte
• spreadsheet-frontend-vue
• spreadsheet-frontend-elm
• spreadsheet-api-go
• spreadsheet-api-erlang/elixir

Related Repos

• spreadsheet-api-tests

Project Setup

• Requires Python 3.8 and pipenv

To install required Python packages navigate to the src/spreadsheetapi directory and run pipenv install.

Running the api server

From the src/spreadsheetapi/service directory run pipenv run uvicorn application:app

Interfacing with the api server

By default the server is available on port 8000 on the loopback address.

Currently the server is set to use the "InMemory" backend. All spreadsheet data created will be cleared between runs of the server with this backend. Switching to the Sqlite backend enables persistence. This is controlled in the config.py file.

FastAPI (the Python web framework used for this process) autogenerates swagger documentation for the API. This is available at http://127.0.0.1:8000/docs. I have not dug deeply into the information presented in this interface, so it may be incorrect or incomplete at this time.

Commit Message Prefixes

Commit messages are prefixed with:

• F: for commits that contain new feature work
• BF: for commits that contain bug fixes
• I: for general code improvements
• ND: for non-deploy commits - these may be documentation updates, changes to CI pipelines, or changes to terraform code

CI/CD and "Production" Setup

This repository uses GitHub actions to automatically test pull requests and commits made to the master branch. The main test suite is located in the external repo spreadsheet-api-tests.

Commits to the master branch are continuously deployed to DigitalOcean's App Platform. Commits can be excluded from deployment by starting the commit message with ND: (Non-deploy commit).

This service uses Terraform to provision and configure the DigitalOcean App Platform on which it is run. App Platform handles building and deploying the Docker container that the service runs within. Deploys are initiated using the DigitalOcean cli tool in the master pipeline.

Development Considerations

Endpoints

Missing Flask's class based views here. Current using classes just as containers for static methods that are functions for each HTTP method the uri implements. Registration is done automatically in the endpoints package __init__.py using a little bit of introspection/reflection..

TODO

• use docker container for service in tests
• remove service directory and shift contents up, will make imports clearer
• config updates - dependent configuration items (eg setting sqlite backend requires db url setting)
• Pre-commit hook for black and mypy
• Redis backend
• Have a base class for Endpoints that is aware of subclasses and registers them with the app when initialized? could be an interesting solution