TobiMayr/url-shortener

A simple urlshortener using django rest framework

url-shortener

Description

A simple urlshortener using django rest framework.

• Receives a URL (originalUrl)
• Generates a unique shortened URL (shortUrl)
• Returns this shortened URL
• Redirects to the original URL when accessing shortened URL
• Tracks visits in database

Running the App

Using Docker

docker build -t urlshortener .
docker run -p 8000:8000 urlshortener
The default setting is production use other settings (development or test) add an argument when building the docker file.
E.g.: docker run -p 8000:8000 urlshortener --build-arg SETTINGS=test

Usage

To shorten a URL a POST request:

Request:

curl --request POST \
  --url http://127.0.0.1:8000/urls/ \
  --header 'Content-Type: application/json' \
  --data '{
	"originalUrl": "https://example.com/long/url/that/would/need/some/shortening"
}'

Response:

{
	"id": "6ffe3537",
	"originalUrl": "https://example.com/long/url/that/would/need/some/shortening",
	"shortUrl": "https://tier.app/6ffe3537"
}

Using a shortened URL to be redirected:

Request:

In develop mode:

curl --request GET \
  --url http://127.0.0.1:8000/6ffe3537

In production, it should look like this:

curl --request GET \
  --url https://tier.app/6ffe3537

Response

The server will return HTTP/1.1 301 Moved Permanently and then redirect to the originalUrl.

Database

The sqlite3 database contains two tables, url and visit,

Tests

Running the tests requires to install every dependency (I used Python 3.9):
sudo pip install virtualenv
virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
python manage.py makemigrations --settings=urlshortener.settings.test
python manage.py migrate --settings=urlshortener.settings.test
./manage.py test --settings=urlshortener.settings.test\

To run coverage

coverage run manage.py test --settings=urlshortener.settings.test
coverage report
Currently it's 98%

FAQ

Why generate a random id (for short url) instead of a hash?

• Two different long urls could create the same hash. Collisions could occur. That's why a unique id is generated.

Why not filter duplicate original URLs to save space in the database?

• It's a tradeoff.
  • Pros: Database size smaller
  • Cons: Complexity of the code slightly increases, checking if the URL exists is an additional operation why would slow down creation of new links
• I chose to allow duplicate original/long URLs. Also, because a more elaborate url-shortener would probably have different users and allow for different TTLs of the same URL.

Ideas for improvement

Code

• add logging
• find a better way to log visits. Right now, only the existing endpoints get tracked. For every new request, code has to be added. Needs to be made more generic
• other fields might be interesting to track. Like latency or status code

Scalability

• create cloud instance for the database (right now db is part of django project)
• run this API in kubernetes (e.g. on AWS EKS). Scaling horizontally, accessing same RDS (which also can be scaled)
• add an in-memory db that is used for the get requests

Functionality

• have user authentication that enables individual users to manage their urls
• let the user choose how long the shortUrl should be valid or set a TTL for every url
• have option to set a short url instead of generating a random one