2

I'm writing a client application that receives a JSON response from a server. In the past I've run into situations where a developer on the server side changes the JSON response in a way that causes the client application to crash. An example of this is when the client expects that a JSON field or subobject will always be present, but a change on the server side causes the JSON to deviate from what is expected possibly returning a null value when null should never be a possible response.

It seem like the server side could always have unit tests that ensure that the JSON response fulfills the contract, but that's susceptible to human error if a developer decides to rewrite a test or simply makes a mistake in testing or misunderstands a requirement. The client side can check that the JSON response is valid, but this would need to occur at runtime and if the server is writing proper tests, the double-checking of the server response by the client would be unnecessary.

Is there a recommended process to ensure that the contract (JSON response format) between the client and server doesn't get broken?

3
  • 6
    Unless you have coordination between the server developers and the client developers, validation on the client side seems inevitable. Jan 20, 2016 at 1:35
  • @RobertHarvey I could imagine a process where the server and client unit tests could both use the same mock objects for testing by keeping the descriptions of mock objects in a shared repository and then having their respective unit testing frameworks pull from that repo and build the mocks, but that seems like it would require a lot of overhead for most projects, but could ensure that unit tests on each end were testing for the same conditions.
    – neonDion
    Jan 20, 2016 at 1:43
  • 2
    The client code should always "test" the server response. Your client code should be built with logic in place to handle bad responses from the server, so that users can easily tell/report when something went wrong. And that logic should be tested.
    – Caleb
    Jan 20, 2016 at 1:46

2 Answers 2

7

Yes, you should validate what you get, but you also need to be a Tolerant Reader:

http://martinfowler.com/bliki/TolerantReader.html

Martin Fowler states that:

My recommendation is to be as tolerant as possible when reading data from a service. If you're consuming an XML file, then only take the elements you need, ignore anything you don't. Furthermore make the minimum assumptions about the structure of the XML you're consuming.

Which, to my experience is a good way to be resilient (tolerant) to changes. Also, if the change is too big to be tolerant, you should consider some strategy to versioning your API:

  1. https://blog.pivotal.io/labs/labs/api-versioning
  2. http://www.mashery.com/blog/ultimate-solution-versioning-rest-apis-content-negotiation

From a more practical point of view, you can also write some test/contracts to your API so that you can continuously check the contracts:

  1. https://github.com/realestate-com-au/pact (see the links at the end of page)
  2. https://github.com/thoughtworks/pacto
1
  • Excellent point. If validating data is to protect against future changes to the server, then it should be handled by passing a version indicator of the server. Otherwise, ensure that only what you require is with the limits of expected server responses and nothing more.
    – Neil
    Jan 20, 2016 at 9:41
4

Why on Earth wouldn't you? A good rule of thumb is to assume that all input is evil until proven otherwise. Blindly assuming that your input will be valid is a good way to wake up to empty databases and overflowing error logs. Just because that input is coming from a (presumably proficient) coworker's server doesn't mean it should get a free pass from you.

At the very least, you want a different error to be logged for Client and Server issues. If the owner of the server changes the response without notice, you want to be able to determine as quickly as possible the source of the error. Aside from covering yourself towards management, it helps to be able to tell at a glance the general area of a problem when your application stops working.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.