DEV Community

Cover image for Best Practices For Designing REST APIs (Part 1)
Osama Riaz
Osama Riaz

Posted on • Updated on

Best Practices For Designing REST APIs (Part 1)

When I started backend development I didn’t pay attention to best practices for developing web APIs and hence my first couple of projects were really a mess. There was no consistency for naming endpoints. I didn’t know how to deal with relationships or how a well-structured representation of a resource looked like. Hopefully, you would not have to go through this phase.

In this article, I’m only going to explain the Resource Representation of an API.

Resource Representation

Firstly, let’s explain what is “Representation” stands for. It could be defined as:

Data that is returned when a client retrieved some resource from the server or sent from client to server.

User can view the representation of the resource in different formats. We call these formats Media Types. All the available media types should encode the same data but just in different format. Out of all the available type, the most widely used one is JSON ( JavaScript Object Notation ). JSON is simple, it’s that simple. This is the reason for it being used widely.


Apart from this, it also has some limitations like a limited data type. It does not support commonly used types like Date and Time and URL. But we can still represent these type using string.

JSON Should Be Simple

Didn’t I just say the JSON is simple? Well, one can make it harder to understand if standards do not meet. JSON is a collection of name/value pairs. The name itself is a string and it corresponds toa property of the resource. It should be kept in mind that a JSON should be self-explanatory. The picture below tells everything.

Notice that in the complex example, ‘user-12348-as13s0-dsdf’ appears on the left of the colon in the position wherein a simple example of JSON I have shown a property name. The JSON name data also does not correspond to a property name

How To Deal With Relations

In almost every project, a resource has some kind of relationships and using simply named properties won’t be that helpful. Therefore, in your response representation, you should consider adding LINKS. Let’s assume that JOHN owns a Dog. Now, there exists a relationship between the owner and its Dog.

Now both representations are valid. In the left, dogID is clearly describing some sort of relationship. But the representation in the right is more expressive and convenient. Why? Let’s assume that you just retrieved the response for John and now you want to get the dog which he owns. If you follow without-links example, you would have to look into the documentation (wish that the docs are up-to-date) to find a suitable endpoint and then construct that in your code to get his Dog(s). But with ‘with links’ example, you already know where to get the response. An increasing number of APIs from major companies on the web are including links in their data like Google and GitHub.

Now it’s maybe tempting to remove the dogID property from the response. Well, it all depends. Links tell us where we are and to where we should go. But there may be some cases when you need dogID other than retrieving dog’s response.

Companies Using Links

Here’s the Google Drive API example:

 "kind": "“drive#file",
 "id": "0B8G-Akr_SmtmaEJneEZLYjBBdWxxxxxxxxxxxxxxxx",
 "etag": "btSRMRFBFi3NMGgScYWZpc9YNCI/MTQzNjU2NDk3OTU3Nw",
 "selfLink": "",
 "webContentLink": "",
 "alternateLink": "",
 "title": "Links.pdf",
 "mimeType": "“application/pdf"
Enter fullscreen mode Exit fullscreen mode

Top comments (0)