grubfinder-server

This is a server for the GrubFinder application, whose purpose is to aggregate all grading and reporting for New York City restaurants and to display them in a digestible (heh) way. You can find the corresponding grubfinder-client code over here.

Development

Dependencies

You will need the following dependencies installed in order to run the application locally:

• Node.js
• MongoDB

Installation

Run the npm install target:

npm install

This will also run postinstall to convert from es2015.

Local Development

Use the start target of npm:

npm start

This will start the local Express server at http://localhost:3000

Running Tests

npm test

Deployment

This project is configured to run on Heroku. See the Heorku documentation for more information.

Dependencies

• Heroku CLI

Git Setup

Run

heroku create

to generate a new Heroku application. This will also add the heroku git remote repository to your configuration:

$ git remote -v
heroku	https://git.heroku.com/grubfinder-api.git (fetch)
heroku	https://git.heroku.com/grubfinder-api.git (push)
origin	git@github.com:benburton/grubfinder-server.git (fetch)
origin	git@github.com:benburton/grubfinder-server.git (push)

Heroku Environment Setup

Add NPM_CONFIG_PRODUCTION=false so that Heroku will install development dependencies as well:

heroku config:set NPM_CONFIG_PRODUCTION=false

Install MongoLab Sandbox plugin

grubfinder-server utilizes MongoDB as its data store. I'm not happy about this, for more than a few reasons. Add the MongoLab sandbox addon to get a MongoDB instance running quickly and (relatively) painlessly:

heroku addons:create mongolab:sandbox

You may need to wait ~30 minutes for an instance to spin up.

Run the NYSDOH Data Import

If you set the MONGODB_URI to the same value as the MONGODB_URI Heroku environment variable, you should be able to run the import from your local machine onto your Heroku instance using the following:

npm run inspection-import

Note that this is particularly slow, as the nature of the data and the speed of import is not friendly to keeping writes off the same records a very long duration of time. As a workaround, I introduced some rudimentary pessimistic locking using async-lock. It's not a pretty solution, but it does get the job done. Or:

Import NYSDOH Data from Backup

I've also made a backup on Dropbox, and included a script to pull it down and migrate into a Heroku environment using mongoimport:

bin/import.sh

Push

You should simply be able to run

git push heroku master

for a working deployment to your Heroku application.

TODOs

Development

• Add a file watcher to hot-reload code changes for easier development

General REST cleanup

• Add Swagger documentation
• Redirect pages for 404s
• Better/clearer 500 handling
• Add routes for processing data import over REST

Testing

• Jasmine unit tests that work async/await keywords for controllers
• Improve overall test coverage
• Load testing

Schema/Database Cleanup

• Move from MongoDB to Postgres. I should have started the project this way 😩
• Remove blank/inconsistent records
• Better data sanitization

Devops

• Consider putting the project into Docker for easier deployment
• Add more verbose logging

Features

• Add geocoding to addresses from NYSDOH so the frontend can use Google Maps