chat-api

Chatting API application in Ruby on Rails and Go

Overview

The API is composed of two separate services:

• Chat API (Rails): Main service which provides most of the core management operations (create, update, get) of applications, chats, and messages, also supports searching through messages in chats using elasticsearch.
• Chat/Message Creation API (Golang): A complementary service to chat API that is responsible for creating chats and messages.

Starting Services

sudo docker-compose down && sudo docker-compose build && sudo docker-compose up

Make sure that docker and docker-compose are installed with dockerd running, also make sure that ports 3000 and ports 8080 are available for the services to run on.

Using Services

Chat API (Rails)

This service exposes these endpoints for operating on applications, chats and messages.

Verb  URI Pattern
----  -----------

GET   /applications/
POST  /applications?name={name}
GET   /applications/{access_token}
PUT   /applications/{access_token}?name={name}

GET   /applications/{access_token}/chats
GET   /applications/{access_token}/chats/{chat_number}

GET   /applications/{access_token}/chats/{chat_number}/messages
GET   /applications/{access_token}/chats/{chat_number}/messages/{message_number}
GET   /applications/{access_token}/chats/{chat_number}/messages/search?keyword={keyword}
PUT   /applications/{access_token}/chats/{chat_number}/messages/{message_number}?body={message_body}

Examples

Creating a new application

$ curl -X POST 'http://localhost:3000/applications?name=app'

# output
{
  "name": "app",
  "access_token": "fPrv7vr57dkUsP4KfZ4BdSmt",
  "created_at": "2019-11-11T19:14:51.589Z",
  "updated_at": "2019-11-11T19:14:51.589Z",
  "chat_count": 0
}

Getting messages

$ curl -X GET 'http://localhost:3000/applications/fPrv7vr57dkUsP4KfZ4BdSmt/chats/1/messages'

# output
[
  {
    "number": 1,
    "body": "Rails stuff",
    "created_at": "2019-11-11T19:18:50.279Z",
    "updated_at": "2019-11-11T19:18:50.279Z"
  },
  {
    "number": 2,
    "body": "Stuff with Rails and Go and stuff with some other stuff",
    "created_at": "2019-11-11T19:18:52.351Z",
    "updated_at": "2019-11-11T19:18:52.351Z"
  }
]

Searching chats

$ curl -X GET 'http://localhost:3000/applications/fPrv7vr57dkUsP4KfZ4BdSmt/chats/1/messages/search?keyword=Go'

# output
[
  {
    "number": 2,
    "body": "Stuff with Rails and Go and stuff with some other stuff",
    "created_at": "2019-11-11T19:18:52.351Z",
    "updated_at": "2019-11-11T19:18:52.351Z"
  }
]

Chat/Message Creation API (Golang)

This service handles creation of chats and messages.

Verb  URI Pattern
----  -----------

POST  /applications/{access_token}/chats/
POST  /applications/{access_token}/chats/{chat_number}/messages?body={message_body}

Examples

Creating a new chat

$ curl -X POST 'http://localhost:8080/applications/fPrv7vr57dkUsP4KfZ4BdSmt/chats'

# output
{
  "number": 1,
  "access_token": "fPrv7vr57dkUsP4KfZ4BdSmt"
}

Sending a new message

$ curl --data '{"body": "Rails stuff"}' -X POST 'http://localhost:8080/applications/fPrv7vr57dkUsP4KfZ4BdSmt/chats/1/messages'

# output
{
  "number":1,
  "chat_number":1,
  "access_token":"fPrv7vr57dkUsP4KfZ4BdSmt"
 }

How This Works

Chat/Message creation API uses Redis for two purposes, first of which is to cache and determine the next chat number and message number to respond to the user with, second of which is to queue jobs to Sidekiq workers hosted at the main Rails API. These Sidekiq workers are responsible of handling requests to create chats and messages in the background, to allow for better scaling in case a huge number of requests is received.

Sending a POST request to create a message follows these procedures:

• Golang Chat/Message creation API first receives this request, gets the application token and the chat number from the request URI.
• It then generates a key that refers to this application token/chat number combination
• It then tries to get the next number using this key from Redis store, if it exists then it atomically gets and increments the value, and if it's not, then it sends a request to the main Rails API to get the current message count.
• After getting the next number, the API queues a job to create this message in Sidekiq, and responds to the user with the number it created.
• When a Sidekiq worker pick up a message creation job, it then adds this message to the messages table, updating a counter_cache automatically.
• Race conditions are handled in both sides:
  • A race condition may occur in the message creation side when two concurrent requests that use the same chat are unable to find a key/value pair in Redis at the same time, when this happens they both send a request to the main Rails API and set the count in store with the same value, leading to two responses with the same number, to handle this redis-lock is used to avoid this issue, requests basically "lock" the key/value pair using this key combination and release this lock after writing to the store.
  • Race conditions are handled at the main Rails side using uniqueness validations on both chat number and message number.

Used elasticsearch to support searching through messages.

TODO

• Add a .env file that contains all the necessary configurations for hosts, ports, passwords, etc. to allow for easier configuration changes, instead of throwing configurations all over the place.
• Add unit tests using Rspec.
• Handle an arbitrary error in the docker image where elasticsearch throws a Java ClassCastException when trying to reindex the messages table, the error doesn't occur outside of the docker container.