RESTful API Designing Guidelines — The Best Practices

Facebook, Google, GitHub, Netflix, and a few other tech giants have given developers and products a chance to consume their data through APIs and became a platform for them.

Even if you are not writing APIs for other developers and products, it is always very healthy for your application to have beautifully crafted APIs.

There is a long debate going on on the internet about the best ways to design the APIs, and it is one of the most nuanced. There are no official guidelines defined for it.

The API is an interface through which many developers interact with the data. A well-designed API is always very easy to use and makes the developer’s life very smooth.

API is the GUI for developers, if it is confusing or not verbose, then the developer will start finding alternatives or stops using it. Developers’ experience is the most important metric to measure the quality of the APIs.

The API is like an artist performing on stage, and its users are the audience.


Terminologies

The following are the most important terms related to REST APIs.

  • Resource is an object or representation of something that has some associated data with it and there can be a set of methods to operate on it. E.g. animals, schools, and employees are resources and delete, add, and update are the operations to be performed on these resources.
  • Collections are a set of resources, e.g. companies is the collection of the company resource.
  • URL (Uniform Resource Locator) is a path through which a resource can be located and actions can be performed on it.

API Endpoint

Let’s write a few APIs for companies that have some employees, to understand more.

/getAllEmployees is an API that will respond with a list of employees. A few more APIs around a company will look like as follows:

  • /addNewEmployee
  • /updateEmployee
  • /deleteEmployee
  • /deleteAllEmployees
  • /promoteEmployee
  • /promoteAllEmployees

And there will be tons of other API endpoints like these for different operations. All of those will contain many redundant actions. Hence, all these API endpoints would be burdensome to maintain when API count increases.

What is wrong?

The URL should only contain resources (nouns), not actions or verbs. The API path /addNewEmployee contains the action addNew along with the resource name Employee.

What is the correct way?

The /companies endpoint is a good example that contains no action. But the question is: “How do we tell the server about the actions to be performed on the companies resource and whether to add, delete, or update?”

This is where the HTTP methods (GETPOSTDELETEPUT), also called verbs, play a role.

The resource should always be plural in the API endpoint and if we want to access one instance of the resource, we can always pass the ID in the URL.

  • The method GET path /companiesshould get the list of all companies.
  • The method GET path /companies/34should get the details of company 34.
  • The method DELETE path /companies/34 should delete company 34.

In a few other use cases, if we have resources under a resource, e.g employees of a company, then a few of the sample API endpoints would be:

  • GET /companies/3/employees should get the list of all employees from company 3.
  • GET /companies/3/employees/45should get the details of employee 45, who belongs to company 3.
  • DELETE /companies/3/employees/45should delete employee 45, who belongs to company 3.
  • POST /companies should create a new company and return the details of the new company created.

Isn’t the API now more precise and consistent?

Conclusion

The paths should contain the plural form of resources and the HTTP method should define the kind of action to be performed on the resource.


HTTP Methods (Verbs)

HTTP has defined a few sets of methods that indicate the type of action to be performed on the resources.

The URL is a sentence where resources are nouns and HTTP methods are verbs.

The important HTTP methods are as follows:

  1. GET method requests data from the resource and should not produce any side-effects. E.g /companies/3/employees returns a list of all employees from company 3.
  2. POST method requests the server to create a resource in the database, mostly when a webform is submitted. E.g /companies/3/employees creates a new employee of company 3.POST is non-idempotent which means that multiple requests will have different effects.
  3. PUT method requests the server to update a resource or create the resource if it doesn’t exist. E.g. /companies/3/employees/john will request the server to update or create if it doesn’t exist, the john resource inthe employees collection under company 3. PUT is idempotent which means multiple requests will have the same effects.
  4. DELETE method requests that the resources, or their instance, are removed from the database. E.g /companies/3/employees/john/ will request the server to delete the john resource from the employees collection under company 3.

There are a few other methods which we will discuss in another post.


HTTP Response Status Codes

When the client raises a request to the server through an API, the client should know the feedback, whether it failed, passed, or the request was wrong.

HTTP status codes are a bunch of standardized codes that have various explanations in various scenarios. The server should always return the right status code.

The following are the important categorizations of HTTP codes:


2xx (Success Category)

These status codes represent that the requested action was received and successfully processed by the server.

  • 200 OK — The standard HTTP response representing success for GET, PUT, or POST.
  • 201 Created — This status code should be returned whenever the new instance is created. E.g. on creating a new instance, using the POST method, should always return a 201 status code.
  • 204 No Content — Represents the request is successfully processed but has not returned any content. DELETE can be a good example of this. The API DELETE /companies/43/employees/2will delete employee 2 and in return, we do not need any data in the response body of the API, as we explicitly asked the system to delete. If there is any error, like if employee 2does not exist in the database, then the response code would not be of 2xx Success Category but around 4xx Client Error category.

3xx (Redirection Category)

  • 304 Not Modified — Indicates that the client has the response already in its cache. And hence, there is no need to transfer the same data again.

4xx (Client Error Category)

These status codes represent that the client has raised a faulty request.

  • 400 Bad Request — Indicates that the request by the client was not processed, as the server could not understand what the client is asking for.
  • 401 Unauthorized — Indicates that the client is not allowed to access resources, and should re-request with the required credentials.
  • 403 Forbidden — Indicates that the request is valid and the client is authenticated, but the client is not allowed access to the page or resource for any reason. E.g. sometimes the authorized client is not allowed to access the directory on the server.
  • 404 Not Found — Indicates that the requested resource is not available now.
  • 410 Gone — Indicates that the requested resource is no longer available which has been intentionally moved.

5xx (Server Error Category)

  • 500 Internal Server Error — Indicates that the request is valid, but the server is totally confused and the server is asked to serve an unexpected condition.
  • 503 Service Unavailable — Indicates that the server is down or unavailable to receive and process the request. Mostly if the server is undergoing maintenance.

Field Name Casing Convention

You can follow any casing convention, but make sure it is consistent across the application. If the request body or response type is JSON then please follow camelCase to maintain consistency.


Searching, Sorting, Filtering, and Pagination

All of these actions are simply the query on one dataset. There will be no new set of APIs to handle these actions. We need to append the query params with the GET method API.

Let’s understand how to implement these actions with a few examples.

  • Sorting — In case the client wants to get the sorted list of companies, the GET /companies endpoint should accept multiple sort params in the query.
    E.g. GET /companies?sort=rank_ascwould sort the companies by rank in ascending order.
  • Filtering — For filtering the dataset, we can pass various options through query params. E.g. GET /companies?category=banking&location=indiawould filter the company’s list data with the company category of banking and where the location is: India.
  • Searching — When searching the company name in the company list, the API endpoint should be GET /companies?search=Digital Mckinsey.
  • Pagination — When the dataset is too large, we divide the data set into smaller chunks, which helps in improving performance and it is easier to handle the response. E.g. GET /companies?page=23 means to get the list of companies on the 23rd page.

If adding many query params in GET methods makes the URI too long, the server may respond with 414 URI Too long HTTP status. In those cases, params can also be passed in the request body of the POST method.


Versioning

When your APIs are being consumed by the world, upgrading the APIs with some breaking change would also lead to breaking the existing products or services using your APIs.

http://api.yourservice.com/v1/companies/34/employees is a good example, which has the version number of the API in the path. If there is any major breaking update, we can name the new set of APIs as v2 or v1.x.x.


Conclusion

These guidelines are based on my experience in development. I would love to know your views on the pointers mentioned above. Please leave a comment, and let me know!

Leave a Reply

Your email address will not be published. Required fields are marked *