My answer to Technicolor's Backend Coding Test.
- MongoDB
- Node.js
npm install
npm test
cp .env.example .env
Then you can configure it to suit your environment.
./index.js
Note, fixture data is loaded when running in development mode.
To see if a user / password pair is authenticated:
curl -XPOST \
-H'Content-Type: application/json' \
-d'{"username":"alice","password":"changeme"}' \
'http://localhost:3000/auth' | json
To retrieve all programmer users sorted by city:
curl 'http://localhost:3000/users?profession=programmer&citysort=-1' | json
To check server health:
curl 'http://localhost:3000/health' | json
To read from a specific dir:
curl 'http://localhost:3000/readdir?path=/tmp' | json
I chose MongoDB because it is easier to work with than RDMS for this specific demo. I don't need to worry about schema at all, and it makes prototyping very easy. I chose Node.js and Javascript because I heard this is what Technicolor uses, and I want to demonstrate that I have good experiece with node.js. I chose Restify as the basis for this API server instead of express or hapi because Restify is a library specifically designed to make restful APIs and it does just enough to get the job done.
API versioning is a feature of routing. With Restify, we can do this easily. Please see Restify API Documentation's Versioned Routes for more details. I also really like the idea of specifying API version via the Accept-Version HTTP header. It makes versioning info orthogonal to other request information, such as query string or url path parameters.
I ran out of time and did not implement pagination. Here is my plan though:
Pagination information should use HTTP Range header. ( Please see Heroku API design. ) This makes the pagination information orthogonal to other query paramters. I would also not use offset / skip, instead I would use lastID. When constructing the query, we filter out any ID that is smaller than lastID and apply a limit to the query. The advantage over offset / skip is that it prevents shifting when new records are added to the previous ranges.
This project uses electrolyte as a Inversion of Control container to achieve Dependency Injection. It's an overkill for this small application, but I wanted to demonstrate how DI can really decouple this applicaiton and making tests easier.
I've worked with Rail's ActiveRecord for many many years, and I wanted to demonstrate a different approach to data storage called Repository pattern, which I think is probably more suitable for JS and functional programming.
You can deploy this on Heroku or build a Docker container or use Puppy / Chef to provision this app. But it is out of the scope of this demo.