Full disclosure, I started off trying to make a TypeScript project with no prior knowledge of Typescript so the file format is what I'd ideally have, with the ts files in src and the compiled (tsc) files in dist. I went ahead and implemented the server is javascript out of time but I'll explain my thought process for the folder structure.
I did create the tsconfig.json file and started writing the controller in typescript before I decided to just go with regular javascript, but they are far from complete.
Ideally, if I wrote the TypeScript, I would separate the source and distribuable files so that I could compile the typescript then run the nodeapp, but for all intents and purposes, we can ignore the src folder and tsconfig.json file.
| Name | Description |
|---|---|
| dist | Contains distributable from TypeScript output |
| dist/controllers | Controllers define functions that respond to various http requests |
| dist/models | Models define Mongoose schemas used in storing and retrieving data from MongoDB |
| dist/routes | Routes define responses to specified http request types |
| node_modules | contains npm dependencies |
| src | Contains typescript files (ideally) |
| db.js | Separate module for the mongoDB connection |
| package.json | Lists out npm dependencies |
| popualate.js | A script I wrote to populate the db with locations.csv |
| server.js | The base server app |
| tsconfig.json | Config settings for compiling server code written in TypeScript |
Because I could use outside packages, I used MLab to create a sandbox MongoDB database and used mongoose to connect to it so that the data could presist between server stoppages. I felt that this was easy enough to implement and a better long term solution since it separates the data management service (database) and the application side services (the node app).
For Find Nearest, I use mongodb's built in $nearSphere geospatial query to easily find the closes shop based on our loc field of [longitude, latitide].
| API Function | URL/Type | Notes |
|---|---|---|
| Create | /shops, POST | Generic Create |
| Read | /shops/:coffeeShopId, GET | Generic Read |
| Update | /shops/:coffeeShopId, PUT | Will update whichever fields given |
| Delete | /shops/:coffeeShopId, DELETE | If you delete a shop that doesn't exist, it'll give back the normal response |
| Find Nearest | /shops/find, POST | Geocodes address given and uses MongoDB built-in $near to find closest shop |
| Find All | /shops, GET | Not specified in the instructions, but an intuitive and simple feature |
I want to note that for Find Nearest, if it's given a address without a city or state, geocoding might recognize it as a nonlocal address. I tried to localize addresses to San Francisco, but an example of an incorrect address is:
https://maps.googleapis.com/maps/api/geocode/json?address=%20270%207th%20St&components=country:US|locality:San+Francisco&key=AIzaSyCawyHgdHUG8d3PKug5ypnGXXmGjKe6kBY
I personally did most of my testing through postman. But for future testing, I would try to use a framework that also support typescript like Jest and using TSLint to catch code quality and style issues.
Packages used and installed through npm
- body-parser - Parses incoming request bodies (req.body) in middleware before my handlers interact with them
- express - Minimal web framework
- mongoose - Interface with MongoDB
- csv - To parse locations.csv
- request - To access Google's Geocoding rest api
- nodemon - Restarts node server when making file changes to make development easier
Typescript, ideally
- npm install -g typescript
- Clone the repository
git clone https://github.com/xurichard/coffeeFinder.git
- Install dependencies
cd coffeeFinder
npm install
- If for some reason the db isn't populated, populate the db from locations.csv
node populate.js
- Run server
npm run dev
or
node server.js
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent