url-shortener

Original task description in README.original.md.

Warning

All of the provided configs & passwords (as seen in docker-compose & Makefile files) should NOT be used "as-is" in production. All of these are simple examples created & committed into this repository to show off basic use cases, and should be used only in local mode for testing purposes.

Usage (showcase)

Available commands:

[sudo] make setup
Sets up the persistent Postgres's Docker volume in .docker directory
[sudo] make server
Starts the entire app (including DB server, backend, frontend & load-balancer).
The app is being served on localhost:8080.
[sudo] make test
Starts backend & frontend test runners in their respective containers.

All commands are marked with optional sudo, indicating that their underlying scripts require higher privileges to run Docker commands (or your user needs to be in the docker group on popular Linux-based distros).

Local dev environment preparation

Ubuntu 20.04 LTS

Install Docker & docker-compose
Install nvm
Install proper version of node (lts/*):
```
nvm install
```
Install lerna globally:
```
npm install -g lerna
```
Bootstrap all services:
```
lerna run bootstrap
```

Services

backend
Main backend service exposing simple REST API. Takes care of the links shortening & persists them in a Postgres DB (currently no other adapters are available).
frontend
Main frontend client served as a web app.
load-balancer
Main gateway, responsible for all traffic routing. Handles the shortened link redirection.

Implementation details

Shortened URLs (slugs) are generated as pseudo-random base64-ish strings of length 10. This gives 2^60 possible slugs, which is approx. 10^18 possible values, making it really unlikely that a collision will occur. That's why there's no link creation retry logic anywhere.
Slugs de-duplication has been left out on purpose. First, it might be desireable by the users to generate different slugs for just one link (eg. for tracking purposes, assuming this project would be extended to support this kind of capability). Second, since the slugs values pool is so huge, there's no need to worry about running out of possible values (excluding malicious usage like DDoS attacks or bots, the security aspect should be handled separately).