Design and Implementation of Coding Platform Using Open-Source Software
Annual Conference of KIPS (ACK 2022)
Summary
The goal was a coding platform that a small team could modify and deploy. The design separates the UI, API, and judge server and reproduces the service setup with Docker Compose.
- Role
- First authoropen-source coding-platform design and implementation research
- Outcome
- Deployed on GCP with problem authoring, browsing, code submission, and judging. Later developed into COU.
A Coding Platform That Small Teams Can Adapt and Deploy
The study combined open-source software (OSS) into an online coding platform that a small team could modify and deploy. Service functions were divided into Docker-based modules to make the platform easier to reconfigure, deploy, and operate.
The design task was to provide problem authoring, submissions, and judging while keeping the scope of each functional change small. Separating the web service, data stores, and judging engine made it possible for a small team to modify the needed component using existing OSS.
Separate Roles for the UI, API, and Judge Server
PostgreSQL, Redis, JudgeServer, and coding-platform each run in a Docker container. JudgeServer grades submitted code and returns execution time and memory usage; coding-platform handles web requests and judging results. PostgreSQL, Redis, and JudgeServer are independent, with coding-platform depending on them. A docker-compose command starts the full service.
Docker Compose defines the containers to run together and their service dependencies. The same configuration can launch the service on a server with Linux, git, and Docker, bringing installation and connection of separate modules into a shared deployment definition.
coding-platform uses Django’s MTV structure with Vue.js for the presentation layer. Django REST Framework provides the API, Django Model handles the database, and the Vue client communicates through axios. The separation lets the front end be replaced with React or Angular, or the back end modified independently.
The UI consumes API requests and responses, while the backend Model handles data access. These boundaries allow changes to a problem-list screen and changes to APIs that read or store its data to be made in their respective layers.
From submitted code to returned judging results
The browser’s Vue client sends a submission through axios to coding-platform’s REST API, which requests judging from JudgeServer. JudgeServer executes the code and returns the result, execution time, and memory usage; the web service delivers those results to the UI.
The layer handling web requests is separate from the layer executing submitted code. Module roles and communication boundaries allow UI or API changes without requiring corresponding changes to the judging engine.
A Working Service from Problem Authoring to Judging
The service deployed on Google Cloud Platform supported authoring problems, browsing the list, submitting source code, and viewing judging results. These operations connected the separate UI, API, and JudgeServer layers into one user workflow.
I conducted the research as an undergraduate first author. This design integrated OSS modules into a coding platform and defined service boundaries and deployment configuration for later changes. It became the starting point for COU.