I'm trying to make a Twitter clone using Go, GraphQL, and Prisma with a Postgresql backend, and a frontend built with Tailwind CSS, Vue, and apollo.
This fork uses session authentication, stores sessions in Redis (port 6420), caches API results (port 6421) on a redis LRU cache, and has an option to subscribe to email notifications for a dweet or user.
The prisma schema is here. This is the single document that will explain to you end to end the way my API functions and stores data.
The redis session store has been configured according to https://redis.io/topics/quickstart, but at port 6420 and 6421. The config and start script files are in the root of this repo.
I use make to make my workflow easier (no pun intended), the commands are:
make run : Run the API, and build the frontend. Not recommended unless really needed. Very bad choice for development, use make api or make serve instead based on your task.
make api : Run the API only
make migrate : Run through only migration of the database. If DB cannot be migrated, you will need to delete it.
make clean : Run through only the deletion of the database. This will delete the database.
make psql : Start the PSQL server
make redis : Start the redis servers
make kill : Kill any processes that are currently accessing the PSQL database. This is only to be used if an error is raised when running make clean
make api : Build the frontend
make serve : Serve only the frontend with hot-reload, auto-refresh, and everything
NOTE: You need ffmpeg added to your path to run this. It uses ffmpeg to generate thumbnails for videos uploaded.
.env contains ACCESS_SECRET, REFRESH_SECRET, DISCORD_CLIENT_ID, DISCORD_CLIENT_SECRET, SENDGRID_API_KEY, SENDGRID_SENDER_EMAIL_ADDR, REDIS_6420_PASS, and REDIS_6421_PASS
cdn_key.json is the key to Google Firebase
This is a part of a uni club induction task.
The tasks assigned to me were:
- Allow users to add a profile picture when registering. They should be able to upload an image as a part of the request. This image must then be stored on a cloud storage service and must be served from a CDN. (Firebase, Cloudinary, etc.)
- Add email verification upon user registration (Only for users registered via email/password).
- Migrate to session-based authentication and store user sessions in a data store like Redis
- Configure your Go Api to use Redis to cache page results. Use LRU for cache updates, and set an expiry time for the cached results.
- Brownie Points: Add a mailing list. Users can choose to receive email notifications for a particular tweet based on tags/user.
- Extra Brownie Points: Add functionality to schedule your tweets. Replace scheduling features provided by email service providers with your implementation using a Message Broker and a Task Queue.
I had 11 days for this task, of which I could utilize only about 7-8 days (I was travelling, and I had to spend time on getting stuff ready to move to campus)
I have finished the first 5/6 of these tasks, the last one I speedran in the last 20 minutes.
This document outlines some of my thoughts on how I implemented the LRU cache on redis.
This is by far my biggest project ever. Yes. Including BruhOS, including the chip8 hardware emulation project I did, even the fullstack Node project I made last year.
This was really fun, especially working with my homie @PseudoCodeNerd / @Madhav Sharma / @mdvsh.
His support and enthusiasm kept me going all through this. He helped me debug some really tough logic issues too. I wouldn't have made any of this without his help and inspiration.
I literally can't imagine myself building the frontend without him. He's really helpful and knowledgeable.
This (hopefully) will be integrated into a bigger Dwitter project later.