Bites

• A platform for discovering restaurants based on ratings, cuisines
• implemented pagination for GET restaurants, GET reviews endpoint
• cached & optimized the GET /:restaurantId/weather endpoint, reduced latency from sub250ms to 20ms on average
• implemented validate() middleware to validate incoming data against strict pre-defined zod schemas.
• implemented early checks for endpoint with :restaurantId to check if the restaurants exists
• implemented standardized responses across all endpoints using successResponse() and errorResponse() helper functions
• implemented centralized error handling middleware()
• Built with Express5, Redis, Zod & Node.js.

Scope of Improvements

• structured unified logging
• api rate limiting
• dockerized deployment (still need to learn docker)

Zod

• zod is used for validating incoming data from users
• working
  • define zod schema
  • infer object types from defined schema
  • use parse() or safeParse() function
  • parse() throws errors if body is not adhering to the defined schema
  • safeParse() doesn't throw errors, it adds a {success: false} field and returns the issues
  • if passed, safeParse() or parse() simply return the data object
  • note : zod will by default allow extra fields in the body and strip them away when it returns back the validated data.
  • that means if I put location: "bangalore" extra field, success won't be false, just the final parsed object won't have location field in it.

import {z} from "zod"

const userSchema = z.object({
  name: z.string().min(1),
  age: z.number().min(1).max(100)
})

type User = z.infer<typeof userSchema>

const user:  User = {
  name: "yash",
  age: 21
}

const result = userSchema.safeParse(user)

if(!result.success){
  console.log("bad request body")
  console.log(result.error.issues)
}else{
  console.log(result.data)
}

• get detailed nested object of errors : result.error.issues
• safeParse() returns

{
  success:true,
  data: T 
}

or 

{
  success: false,
  error: zodError 
} 

• issue arrays lives inside the zodError. so error.issues is what's the meaning full data at.

Redis

• cache aside pattern
• while reading - 1st read cache, if cache misses -> read from primary database.
• while writing - write to both cache with expiry and primary database asynchronously.
• redis has multiple datastructures, few of them are
  • redis string : a simple key:value pair
  • redis list : a linked list of string with O(1) top and last retrieval, can be used as a normal array like datastructure with O(1) retrieval with indices. (0,1,2,3,4...)
  • redis json : to store json objects linked to a key
  • redis hash : to store hased flat objects, instead of storing user.name, user.email separarely, we hash the whole user object against a single key.
  • redis set : unordered collection of unique strings, can be used as indexed database.
  • redis sorted set : collection of unique strings stored in ascending order. can be used for leaderboards, recent ratings, recent data.
  • redis bloom filters : probabilitic data structure like set, used to test whether an element is a member or not, used to implement ultra quick "username is taken" checkes. can be used to avoid duplicacy of data and fast duplicacy checks.

Redis Server Setup

• npm install redis : redis is client-side driver package and needs to connect with local or cloud redis server.
• redis connection string looks like : redis://default:@.io:13487 (taken from official redis cloud platform)
• by default local redis server serves on PORT 6379.