README

Implementation Overview

Flask is used as the backend framework because of its lightweight-ness and customizability for api servers.

mongodb is used as the db because we are dealing with unstructured data.

docker-compose is used for easy setup.

The logic for merging hotel details from different sources is in Models/HotelDetail.py, where the class takes a parameter of available results, then modularizes the addition of individual detail in the form of a class method. This way, they can be modified/added/removed independently of one another.

Setup

Navigate to a desired folder and clone this repo.

git clone https://github.com/hesingon/ascenda_assignment.git

Go to the repo directory, build and run the project with docker-compose.

cd ascenda_assignment
docker-compose build
docker-compose up

At this point, the api should be accessible at localhost:5000/find_hotels

Either hotel_id or destination_id can be passed in a parameter. e.g.

http://localhost:5000/find_hotels?hotel_id=iJhz

http://localhost:5000/find_hotels?destination_id=5432

Note that the two should not be passed together. If they are both present, the server will only handle hotel_id

Configuration & Settings

You can do a few configurations by changing configs/settings.py.

HOTEL_UPDATE_INTERVAL represents, when server receives a request for a single hotel, how much time in minutes the server should check the supplier urls again for any updates of the hotel.

DESTINATION_UPDATE_INTERVAL similarly represents the time in minutes for updating all hotels in the same destination.

SUPPLIERS_ENDPOINTS are the list of supplier endpoints from which the server will check for updates of hotels periodically.

Design Consideration for Performance

Whenever server receives a request, it queries db first. If db does not contain info for that hotel, or it realized that the hotel was last updated a long time ago, it will query all the supplier endpoints to get and merge the latest info. We check for updates for a configurable interval (e.g. every 3 hours or 1 day).
The querying of multiple supplier APIs is done in a multi-threaded manner with aiohttp to minimise request latency.

More possible improvement

We can include redis as an in-memory db on top of mongodb that allows fast retrieval of the more frequently read destinations.

hesingon / ascenda_assignment