If there's one thing developers have to deal with most of the days, it's APIs. For web developers, this often means dealing with HTTP and external services. The question is, how to test the APIs and make sure everything works as we expect.
For this reason, I am interviewing Honza Javorek, the author of Dredd, a tool designed for this particular purpose.
Since 2011 I have been helping to drive the growth and success of the Czech Python User Group. For years I have been participating in volunteer-driven meetups, courses, workshops, conferences, nonprofit, and more.
I have the privilege of working on Dredd, an open source project, as my day job.
I'll assume the reader knows what web APIs are. Dredd is a command-line tool which tests whether a web API works as expected according to its specification.
When you are about to develop an API, you usually start with an idea of how it should work. APIs are interfaces between systems, teams or companies so to avoid future problems it often makes sense to make the ideation process collaborative. Ideally, to share and discuss the design of the API with other people, you write it down into a document.
While a plain text file would be enough for collaboration, if you describe your API in a popular, structured format like API Blueprint or OpenAPI you get the benefit of having a machine-readable source of truth for your API, which can be then used by various tools. For instance, you can automatically generate documentation from such a file.
FORMAT: 1A
# My API
Below is one of the most straightforward APIs written in the **API Blueprint**. You can use [Markdown](https://daringfireball.net/projects/markdown/) here.
# GET /users
+ Response 200 (application/json)
[
{
"id": 1,
"name": "@survivejs",
},
{
"id": 2,
"name": "@honzajavorek",
}
]
When it comes to developing the API server implementation, Dredd can read the API description document, learn about the expected behavior of the API, then make requests to that API and confirm that it has returned the expected responses:
dredd api-description.yaml http://127.0.0.1:8000
Dredd ensures that the API implementation never gets out of sync with the specification so your team and your users can be sure that the essentials you promised in the documentation are always correct.
Moreover, since Dredd is an open source command-line tool installable from npm, it can be easily used as part of the API project's test suite and regularly used during development. It can be run on Continuous Integration such as Travis CI or Jenkins to ensure that no changes to the project can break the contract given by the API description document.
I don't think there's a real competitor to what Dredd does except for custom-made solutions.
It's important to understand that Dredd doesn't completely replace other automated tests. You should still unit test your code and have integration or end-to-end tests to cover various corner cases. Dredd helps you to ensure the essentials you documented and promised in your API description file will always work in the implementation.
Dredd is a project developed by the company I work for, Apiary (acquired by Oracle in 2017). Apiary pioneered the market of API design tools and contributed a bunch of open source projects to the world on the way, such as Dredd, API Blueprint, or API Elements. The original author of Dredd is Adam Kliment. I took over the project in 2016, leading and maintaining it until today.
And why did we create Dredd? It clicks together with the design-first approach to developing APIs which we've been preaching for years.
Especially people focusing on the frontend and working on API-backed SPAs know the frustration experienced when they need to work with an API which has been designed without their needs in mind. It is a common problem, and I believe it's mostly this frustration which drives people to try GraphQL instead of web APIs based on embracing HTTP and some of the ideas behind the REST architectural pattern.
In Apiary we firmly believe that hard-to-use APIs come to this world mainly by being developed without any communication with their future users and by documentation being generated out of the final code instead of written with human beings in mind. The right way to develop APIs, according to us, is to design them first, to write the documentation first, and to communicate before any code gets written.
But if you're supposed to write down a document and only then start to write code, how do you ensure that the implementation is done according to that document? The answer is Dredd. Dredd is essential for the design-first approach. So to be able to preach design-first, we had to create Dredd.
We've introduced experimental OpenAPI 3 support in Dredd and are currently working on making it a first-class citizen among the supported formats.
Except for that, I have many ideas on how to improve the developer experience with Dredd. The reporters could have better arranged and more useful output, for example.
As I already mentioned, I can see people trying GraphQL as an alternative to web APIs. I believe there are good reasons why those two should coexist in symbiosis instead of replacing each other. I think that the future of Dredd in such a world isn't in supporting GraphQL, but in doing more for the web APIs so that developing them feels easier.
Dredd could, for example, introduce a way to test not only the API server implementation but also the API client so the contract would be validated from both sides. Imagine Dredd could read the API description document, then test your frontend or mobile app requests, and in the end give you a verdict on whether they conformed to the API description.
One more thing I'd love to have in Dredd is testing scenarios, i.e., series of requests and responses.
Even if you don't build web APIs in particular, I believe the design-first approach is a general idea you can apply to any interface as a programmer, whether it is an interface of a library you're building or a user interface of your website.
My advice? Read Readme Driven Development by Tom Preston-Werner and How I Develop Things and Why by Kenneth Reitz. Then embrace the concepts and build stuff other people will love to use.
You should interview Mila Votradovec (@MilaVot) from Snyk. The company is doing a great job allowing people to use the open source ecosystem of their language to the maximum while making it easy for mere mortals to pay attention to security.
Thanks for the interview! I'm happy to answer any questions about Dredd on Twitter or in Dredd's issues.
If by any chance anyone reading this happens to be interested in Python as well, consider visiting this year's edition of the PyCon CZ happening on June 14–16th. It's the first time it's in Ostrava, and the team has managed to secure a fantastic venue for the event - former steelworks and coal mines premises.
Thanks for the interview, Honza! For me, it's an entirely new way of testing I hadn't considered before and I can see value in the approach. I'm sure other people will too.
To learn more, check out Dredd site and Dredd on GitHub.