dfdeshom / test-range

RangeDB

RangeDB is a toy in-memory threadsafe query engine that users can interact with via an HTTP server. The objects it accepts looks like:

{
 "identifier": "foo",
 "ranges": [[12,34],[37,440],[460,800]]
}

identifer and ranges mut be present.

Operations

There are 3 kinds of operations this server accepts:

• /write which is POST operation that writes a new range object to the database. If the object already exists, it will be overwritten. The POST data mut be in json string format like so: {"identifier":"data", "ranges":["1","4"]}
• /get?identifer=<id> which retrieves a range object that matches idetifier.
• /ranges?range=<range> which gets a list of objects that overlap a specific range. <range> is of the form [start,end].
• / which lists all objects in the database

Building/running

To run:

$ pip install -r requirements.txt
$ ./runserver.sh

Assumptions and implementation notes

We use a dict as the underlying data store, for which most atomic operations are safe (get, set). For operations that do a get-and-set, we use a reentrant lock.

Although the code is threadsafe, we assume that the server will be deployed in a multiprocess environment. This is because meaningful operations like finding a range are CPU rather than I/O-bound and use of threads in Python is prescribed for operations that are I/O-bound. This also doesn't make evented web servers like Tornado ideal for use, which is why we chose the Flask web server as a test server implementation.

A multiprocess environemnt also implies that clients would have to be smart about which servers they write and request data to. Such client is not included here.