Roomba Puzzle

I've been running into some problems recently; I bought a Roomba a few months ago, and although I know the exact dimensions of my room, and the locations of dirt at any one moment (I track it religiously), I simply can't figure out how many patches of dirt would have been swept up if I give it a certain set of directions! I'd also like to persist my Roomba's journeys to a DB, so that I can go back and remember the good times.

Sounds like a bit like a programming exercise, doesn't it? https://gist.github.com/alirussell/9a519e07128b7eafcb50

The Solution

This problem boils down to a move-and-record - we want to extract relevant information from a JSON payload, walk through the instruction set, updating our location recording any coordinates we visited that contained dirt.

I've implemented this as a Flask app in Python 3.7 + Redis, which is tested using PyTest, and Dockerised. There were a few interesting points in this exercise:

We need to clean the starting position, as the Roomba is always on.
The return values should be the number of patches cleaned, not the number remaining.
When we exceed the room dimensions, we don't move, we hit the wall for as many directions as we don't progress.
If we're persisting to a DB, and this is a functional problem, we can cache responses instead of re-executing. We need to index on a composite key of all inputs.
This problem would lend itself to recursion pretty well

Getting Started

To run the Docker image, simply run the following command:

docker-compose up -d

This spins up our app server running on Flask, and a Redis server - exposing the REST API on port 80. You can perform requests in a REST client or programatically using the routes below.

To run the test suite once the service is up, in the root directory:

pytest

Routes

POST /

Headers

Content-Type: application/json

Body

{"roomSize" : [5, 5], "coords" : [1, 2], "patches" : [[1, 0], [2, 2], [2, 3]], "instructions" : "NNESEESWNWW"}

roomSize : A two-element list of the integeric size of the room in the format [x,y]. Defines the maximum bounds for our Roomba.
coords : A two-element list of the starting position of the Roomba, in the format [x,y].
patches : A list of lists, two-elements each, integeric. These are the coordinates where we know there to be dirt. Each time the Roomba passes over one of these, we count it as cleaned.
instructions : a stringified list of North, East, South, and West instructions, determining where the Roomba will travel.

Response

200 - Successful {"coords" : [1, 3], "patches" : 1}

coords : The ending position of the Roomba
patches : The number of patches we were able to clean in this run

400 - Bad Request

The JSON POSTed was in an incorrect format, we couldn't proceed.

Expansion

Following on, there's a few ways I would expand this project with more time, or if productionising in some way:

We're only logging to STDOUT if the FLASK_ENV is 'dev'; we probably want to add some proper logging infrastructure.
Ideally, we might like a frontend to display Roomba progress - I'd tackle this in Javascript with a roomSize grid of divs, AJAX calls to Flask and an update of background-color. As the task sort of demands that the algo executes in a single step, I didn't want to rewrite it in JS which is probably what I'd do as it's not a complex task, and there are no server-side components. (JS is not an accepted solution language)
For the sake of keeping things easy, I've indexed on a SHA256 of the inputs, if we were concerned about bigger rooms and larger instruction sets, we could try and calculate whether a certain set of instructions has been executed in a room that had smaller dimensions and didn't hit the walls, meaning we could return a cached response even though the roomSize is different.
Simarlarly, we could optimise by not even running the algo if there are no instructions, and not keeping track of patches if none were provided.
If # of Roombas > 1, that would get interesting. Each device would run in parallel and sync patches of dirt that had been cleaned, as well as checking whether the next position already has a device in it, or if it's about to.
Roombas have to return to base to charge up (I believe) - it'd be interesting to keep track of battery level and have to navigate back to 0,0 if it got too low, cleaning on the way.

psedge / roomba