A small application for the bike parking Uber Web Coding Challenge.

Bike parking data (mostly just coordinates) was thrown into a Mongo database. I expose two endpoints: /index to render the page and /search to perform AJAX search requests. Searches return parking spots within 250 meters of the given lat/lng. I don't show 'IN PROGRESS' spots to the user since those aren't useful.

You could call this a full-stack solution, in the sense that I spent some time on the frontend. I wouldn't call myself a frontend developer, though; I only know enough to be dangerous.

I considered implementing this as a static site (with all data loaded in-memory) but I didn't want to bother with implementing querying, so I went with Mongo as a data store because it's web scale. Just kidding. It has built-in geospatial querying and it's easy to setup.

I used a small starter template for Flask that I made to give me some initial directory structure and such. I actually ended up rewriting a bunch of it (to add testing with pytest, tox support, flast-assets) so I'm excited to be able to merge that upstream. Flask is nice for getting something out the door quickly.

I wanted to demonstrate familiarity with some tools without going too overboard. Notably I don't use any frontend frameworks like Bootstrap or jQuery. By the time I added Google Maps support I realized that I had almost all of the vanilla JS I needed, and I was able to fake a responsive layout with a few lines of CSS. The app is simple enough and already has enough frills that adding these would just be bloat and not a good use of time.

Colander is used to validate user input on searches.

Never worked with Google Maps before, so that API was new to me. It went OK.

Some stuff I would work on or do differently if I had more time:

• The other two problems -- film locations and food trucks -- are similar to this one (show me something nearby). Those data sets were missing lat/long coordinates, though, and I didn't want to deal with geocoding. If I had time to geocode that data it would be really easy to drop it in and display in addition to the bike data.
• Javascript is not tested. I hear good things about PhantomJS.
• IE is probably not supported at all. This doesn't bother me, but it would need to be addressed eventually.
• Rotating the screen on mobile hides the map, which sucks.
• Integration with Travis CI.
• I wanted to play with Ming for testing Mongo support.
• I'm not using argparse or anything in my main app.
• I'm not using the SODA API so my data is going to become stale. If I kept stuff in Mongo I would would want a process to periodically sync it.
• The frontend could be richer -- details about parking spots like how long it's been around. The usefulness of that is questionable, though.
• API errors on /search don't have any useful debugging information. I would want to make those return JSON.
• Inline comments are a little lacking because I spent more time on this doc. In practice I would want to generate Sphinx docs.
• I'm missing a favicon.

Running the application locally requires Mongodb 2.4.x:

$ brew install mongodb

Python dependencies are handled with virtualenv. You can install these in your own environment

$ pip install -rrequirements.txt

or, if you have tox available, you can use its environment:

$ make test && source .tox/test/bin/activate

You will need a Google Maps API -- place that in an api_key file at the root of your repository.

I haven't actually verified that this packaging works, so good luck!

$ make coverage
...
$ make server
--------------------------------------------------------------------------------
 * Running on http://0.0.0.0:8080/
 * Restarting with reloader
--------------------------------------------------------------------------------