go-with-vue

Simple web app build with Golang, PostgreSQL, Docker and Vue.js

Getting started

First, we have to create Docker images of components by executing docker-compose build. Afterward, we execute docker-compose up and this involves 3 stages:

1. It spins up a PostgreSQL instance container that runs on port 5432
2. After the database instance is ready, a small Golang program named newsfeeder gets executed. It fetches RSS feeds from NewYork Times and insert the entries into the database.
3. Lastly, a Golang web server gets spun up after the database and newsfeeder are ready.

Main front page

By accessing to http://localhost:8080/, we see a list of news articles that were fetched from NewYork Times RSS feed along with Rating for each article which is implemented for this app.

Top 5 best rated articles

By accessing to http://localhost:8080/bestnews, we see a JSON object that contains an array of the five best rated news articles. This object is updated every 5 minutes, so accessing the endpoint does not make the back-end hit the database, therefore reducing the database load. Initially, all of the articles are not rated, therefore we see an empty array for the first 5 minutes.

Architectural Decision

One of the common design problems when developing a web application with Golang is how to manage a database connection such that the application runs efficiently and is unit-testable. In order to solve these issues, we use a context object and Golang's interface feature for our database type. In api/main.go we can see a struct:

type Context struct {
	DB       database.Database
	BestNews *models.BestNews
}

that holds a database object that implements Database interface and a bestnews object that gets updated every 5 minutes. The Context object gets passed to the handlers for endpoints and each handler can have an access to the database object. This brings us the following benefits:

1. Eliminating the use of global variable for a database connection object.
2. Each handler does not have to instantiate a new database object, which is computationally expensive.
3. Making the handlers unit-testable by implementing a mock database object that implements the database interface.