We built a social network!
This academic year we were asked to build a web app as a group project for the database course, we could have chosen between an e-commerce or a social network.
My group and I decided to stick with the social network for the freedom it would give us.
The rules were simple: “the web app must use flask, sqlalchemy and postgresql”, and so we did, but we were given absolute freedom in every other aspect of the project.
Since the project is already well documented here, on this page I will only discuss some of our thought processes and implementation choices.
Quick tour
Here’s a brief run-down of what bonfire looks like from the users and advertisers point of views.
Users
The followings are the main pages of the social network, namely the explore section and profile section.
Everyone can explore bonfire, but only those with an account can interact with it, the registered users can in fact create and interact with posts and other users.
Advertisers
The social network in addition to letting users interact with each other, must’ve had a focus for advertisers, so that they could publish targeted ads, below are the advertisers main pages.
It is possible to create several ad campaigns, each one with its own ads, and then it is possible to monitor their performance through a dedicated dashboard which displays: impressions, ad readings and clicks over time.
RESTful API
We decided to exchange information between the backend and frontend through json messages instead of using flask’s built in functions, this brought us several advantages, including:
- The ability to upload content to the frontend without refreshing the page
- Facilitated parallel development with git
- Early testing of APIs testing thanks to tools like postman
Each API call is an atomic database transaction which uses ORM for querying the database, in this way the codebase is also portable to other DBMSs.
Interests updates
Interests are represented by tags, which are limited and chosen by the app owner.
When creating a post, multiple tags can be included, and when another user interacts with it, the user’s interests are changed according the tags in the post:
- Increasing the interest for the included tags
- Decreasing a bit the interest for non included tags
In this way we can always suggest an ad based on the user’s trending interest.
Ads recommendation
With each chunk of posts the server sends to the frontend, it also sends an ad, which in order to be chosen must belong to an ad campaign that has enough budget to pay the expected fees and that targets the user’s trending interest.
Once the ad campaign is chosen the ad is picked based on the percentage of appearance selected by the advertiser.