Currency Conversion Service Case Study for Arf

This document provides useful informations about the project and its deployment process.

Table of Contents

1. About the project
2. How to start
   1. Run repo via git-clone
   2. Test sample endpoints
   3. Swagger docs
3. Business Logic
4. System Design
   1. Data models
   2. API Layers

About the project

This project provides services which users are able to login with their credentials and convert the available currencies they want to. Also users have multiple wallets. Currency rates are static at the moment and defined in a config file.

How to start

Run repo via git-clone

First provide an app.env file. Use the sample.env format. You can copy or change the name of sample.dev.env.

System user created initialy as provided in the app.env file. It has all the wallets with sufficient balances.

To test the system, create a user & wallets with different currencies.

Use docker exec -it postgres psql -U ${USER} -d ${DB} -W to access the db. So you can add balance to the users & modify records.

Run make build to build the project with docker-compose, then make up to run it.

Sample endpoints

I tested endpoints in Postman, so here you can see the sample requests.

### Swagger docs

Hit {BASE_URL}/swagger/index.html to see the generated swagger endpoint list.

System Design

Data Models

Basic data blocks used in the system:

API Layers: Offer service

Offer service implemented as a 2-step process which requires an approval after it.

Endpoints are isolated to separate DB domains.

Business Logic

The main point of the conversion is creating Offer, which has the required conversion info. rate is tricky since I assumed that frontend fetches the rate from db & use it to request an offer. Then we check if it's same with the one in our system.

How to determine rates according to the markup rate

I used a simple logic as below:

e.g

from: USD
to: TRY
amount: 100
rate: 10
markupRate: 0.1
rate * (1 - markupRate) = 9

system should have at least (100) TRY
user should have at least (100/9) = 11.1 USD
(it's supposed to be 10 USD if no markup applied)

Basically, instead of changing the currency rate, I charged user more according to the markup rate. So user pays more to get the desired amount. There could be alternatives but I thought it would still make us profitable & also practical to implement.

What about freezing the currency rate?

I used Offer object with a timestamp. System requests the approval of this offer. Since Offer has a rate and timestamp in it, I check the timestamp difference when approve request hit. If 3+ minutes have passed, I reject that offer by responding an offer timeout.

What about tests

I really tried, but couldn't cover all of them due to the lack of time & some platform issues. I still think I could cover most of them over a wider period of time since I implemented the test suits, mocks & stubs.

Cheers ^^