Opendoor Take-Home Coding Challenge

This API was written as a part of the Opendoor interview process. A full problem description can be found in PROBLEM.md

Querying

Data is loaded from https://s3.amazonaws.com/opendoor-problems/listings.csv on boot.

The available endpoints are:

Endpoint	Supported Methods	Description
/	GET	Returns the contents PROBLEM.md.
/listings	GET	Fetch a JSON list of house listings.

The /listings endpoint supports the following query parameters:

Parameter	Description
min_price	The minimum listing price in dollars (inclusive).
max_price	The maximum listing price in dollars (inclusive).
min_bed	The minimum number of bedrooms (inclusive).
max_bed	The maximum number of bedrooms (inclusive).
min_bath	The minimum number of bathrooms (inclusive).
max_bath	The maximum number of bathrooms (inclusive).

Requirements

• Python 2.7 (currently, MySQL-python is not Python 3 compatible)
• VirtualEnv
• Foreman

Running

The easiest way to run the app is to simply deploy it to Heroku. View the Deploying section for an easy, one click button to do this.

To run the app locally, start by creating a .env file containing a CLEARDB_DATABASE_URL and your credentials:

CLEARDB_DATABASE_URL=mysql://username:password@hostname.com/database

Next, run the following commands to install the Python dependencies

virtualenv env
source env/bin/activate
pip install -r requirements
foreman start

If you are running the app on OSX, you may encounter an error while installing MySQL-python. This can generally be cleared up by installing the MySQL C dependencies with brew install mysql.

After running foreman start, the app should be running on port 5000.

Testing

Currently, there are no tests (see the Future Improvements section for more info).

To run the Python linter, run:

make lint

Deploying

This project is designed to be run on Heroku with a ClearDB MySQL backend (the free Ignite tier provides enough storage for the Opendoor CSV file). Information on manually deploying to Heroku can be found here.

Requirements Breakdown

At a minimum:

• Your API endpoint URL is /listings.
• Your API responds with valid GeoJSON (you can check the output using http://geojson.io).
  • The output of this app has been linted against GeoJSONlint.com.
• Your API should correctly filter any combination of API parameters.
• Use a datastore.
  • This projected is backed my MySQL. The Heroku deploy button creates a ClearDB MySQL instance for storing and manipulating the data.

Bonus Points:

• Pagination via web linking (http://tools.ietf.org/html/rfc5988).
  • The Link header can be easily viewed with curl by running curl -i "http://API-URL:5000/listings?page=30". first and last paging is always available, and prev and next are available when applicable.

Future Improvements

Given more time, I would like to make the following improvements:

• Tests (primarily unit and acceptance).
• Filtering based on the status of the listings. Currently, all listings are being returned with no indicator of what state they are in (active, pending, or sold).
• Currently, every query against /listings creates a new MySQL connection. Ideally, one connection could be created when the app launches and be refreshed when a disconnect is detected.
• Page sizes are currently static (locked to 15). There should be a count parameter to make this configurable.
• Better exception handling. There area a couple places where exception handling could be improved, such as around the database connection and around the handling of the CSV data.
• Sorting. The /listings endpoint currently always sorts by id.