Katkat
Back

Design automated regression tests for an API

TikTokSoftware Engineer · Technical Screen
EasySoftware EngineerTechnical ScreenSoftware Engineering Fundamentals

Last updated

This question assesses the ability to structure a comprehensive automated regression suite for a RESTful API, focusing on test case selection, validation strategies, and maintenance.

Problem Statement

Imagine you are maintaining a RESTful API for a user profile service. The service has three primary endpoints:

  • GET /users/{id}: Returns profile details.
  • POST /users: Creates a new user.
  • PATCH /users/{id}: Updates user fields.

The service is undergoing frequent feature updates. Your task is to design an automated regression testing strategy to ensure that new deployments do not break existing functionality.

Requirements

  1. Define the core components of your test suite (e.g., unit, integration, end-to-end).
  2. Propose a mechanism to validate the schema and data consistency across different API versions.
  3. Explain how you would manage test data in an isolated environment to ensure repeatability.
  4. Define criteria for what constitutes a "passing" test suite before a deployment can proceed.

Constraints

  • The tests must be repeatable in a CI/CD pipeline.
  • You must account for external dependency mocking (e.g., databases or third-party auth services).
  • Provide a brief example of how you would assert a successful POST /users request and verify that the subsequent GET /users/{id} contains the expected data.

Hints

Hint 1

Focus on the 'testing pyramid' and distinguish between testing isolated logic versus testing the full request-response cycle.

Hint 2

Consider how you would handle 'tear-down' logic to ensure that test data created during a run does not pollute subsequent tests.

Hint 3

Think about how contract testing can help detect breaking changes early before running expensive end-to-end flows.

Follow-up questions

  • How would you handle asynchronous API operations or webhooks in your regression suite?
  • If your test suite takes 45 minutes to run, what strategies would you use to reduce this time without sacrificing safety?
  • How would you distinguish between a 'flaky test' and a legitimate regression bug, and what process would you put in place to handle them?