A simple but scalable ShortURL service written in Golang.

• server
  • API backend server
  • Includs a token collection service that collects unused ids before getting request
  • May have multiple server running at the same time
• cleaner
  • Small service that periodically clean up the expired records from database.

Run docker compose up or docker-compose up, and voila! The service will be served on localhost:8000.

go build cmd/server/server.go
go build cmd/cleaner/cleaner.go

Configurations are loaded from the environment variables, if not provided, then the default values will be used.

• LISTEN_ADDR: The address that the server listens on, defaults to :8000
• BASE_URL: The base URL of the service, defaults to http://localhost:8000
• MAX_TOKEN: The maximum amount of tokens to be generated and stored offline, defaults to 5000
• MAX_URL_LEN: The length limit of url string, defaults to 1024
• MAX_ALIVE_DURATION: The maximum duration the record stays in database, defaults to 8760h
• MONGODB_URI: The uri to connect to mongodb, defaults to mongodb://localhost:27017
• DATABASE: The database to use, defaults to shorturl
• RECORD_COLLECTION: The collection to store records, defaults to records
• MEMCACHED_ADDRS: The addresses of memcached services, separated by comma, defaults to localhost:11211

Example Request Body (application/json)

{
    "url": "https://www.dcard.tw/f",
    "expireAt": "2021-02-08T09:20:41Z"
}

Response

{
    "id": "8RjJue",
    "shortUrl": "http://localhost:8000/8RjJue"
}

Go package echo is adopted as the web framework for its simplicity, extensibility and high performance. Other golang web framework packages like gin were also considered, but I have experiences with developing with echo, so it's adopted in final.

Considering the use case, we basically only need to do key-value storing and searching instead of complex relation quering. So in this case, MongoDB seems to be a good choice. Also, MongoDB is well-known for its distributed architecture and high scalability, while SQL-based database like PostgreSQL is more diffcult to scale horizontally. Therefore MongoDB is adopted as the database.

Go package mongo is the MongoDB Driver API package maintained by the MongoDB official, we use it to connect and operate on our database.

Memcached is a cache service that can be distributed deployed, in this project Memcached is adopted as the cache for speeding up the resolve time for frequently requested shorturls.

Go package gomemcache is used as the interface to interact with Memcached server. Since there is no official package released, the most imported package is choosed.