Dzhavat Ushev

Making HTTP requests inside Azure Functions

Published: July 9, 2019 | Updated: April 26, 2020

I have an idea for a feature that I want ot add to this blog. It’s something I wanted to try but never found the time. That’s probably how most ideas are. Anyway, this past weekend I finally found a couple of hours.

The feature is something I’m going to implement soon. It’s already implemented! Check out the front page. It’s about showing some information from a third-party service. In order to call the service, I have to make an HTTP request and include an access tokens. It’s not safe doing that directly from the front-end since the tokens are supposed to be private. What I need to do instead is expose my own endpoint on a server and call the service from there.

My first choice was to use Node.js. I made it work locally but then realized that I had to host it somewhere, make sure it’s running, etc. It was too much hasle. At this point I decided to give Azure Functions a try. They seemed to be a good fit for this case. I hadn’t tried them before so it was also an opportunity to experiment with something new. Yay!

I signed up for Azure and in less than 5 min everything was up and running. Except that the HTTP request to the third-party didn’t work! Then I spent the good couple of hours trying to figure out why. Eventually, I “complained” about it on Twitter and, with some help, managed to make it work.

Here’s an example of the code I ended up with. I’ll go through some key points underneath:

const fetch = require("node-fetch"); // 1

module.exports = async function (context, req) { // 2
    const accessToken = '...';

    const url = 'https://api.github.com/user';
    const headers = {
        'Authorization': `token ${accessToken}`
    };

    await fetch(url, { headers }) // 3
        .then(response => response.json())
        .then(response => context.res.json(response)); // 4
};

1. I’m using node-fetch to make the request. The reason for this is because it has an API similar to the fetch that runs in the browser. You can use whatever other package you like or even the native http(s) implementation in Node.js.
2. It’s important to note that the exported function is marked as async. This was something I overlooked initially and that was probably the reason for wasting some precious time. Having the async in there means that you don’t have to call context.done() to indicate that the function has completed. This happens implicitly. I found a lot of code examples on the Internet that don’t declare the export function as async. In this case, context.done() must be called. Generating Azure Functions v2.x automatically declares them as async.
3. Using the async/await combination is a recommended pattern. Another alternative would’ve been to return the Promise.
4. context.res.json(...) is a nice method for setting both the Content-Type to application/json and body to whatever the response is at the same time.

This was my first experience working with HTTP triggered Azure Functions. Looking back, it’s really not that complicated. I could’ve had my endpoint ready in a few minutes. Now I’m struggling with Timer triggered functions but that’s a subject for another post.

Share on Twitter