Different breeds of automated testing, for comparative purposes.
Forks, requests and PRs welcome. Please reuse under creative commons (see footer.)
Test correctness at the following sizes:
Smallest: test code corresponds closest to testable code; other layers injected or mocked.
Medium: testable code involves more than one layer or unit; comined behaviour tested.
Largest: entire application invoked for each test.
(aka "end-to-end testing", but describing the full extent of the code stack, not the full extent of the user journey in Behavioural testing below.)
Confirm testable code works successfully, by defining success as:
Testing of code's behaviours, treating the codebase as a black box.
*(aka "black-box testing".)
*(Behavio(u)r-driven development is broadly a behavioural implementation of Unit testing, carried out within the methodology of test-driven development.)
Testing of code's functionality, often with one eye to the internals and how they might fail.
Testing only the end result of many different actions, possibly including external behaviours.
Write the test in order to satisfy the following business requirement(s):
Prevent a bug from arising during future development.
Confirm the current state of development works as required.
Determine the capability of the testable code to withstand heavy traffic.
Preface more granular (and possibly slower) testing, by looking for simple but severe failures only, possibly using a subset of the overall test suite.
(Sometimes used instead to describe a subset of Performance testing, when the traffic is increased until the code "catches fire.")
Different tools—to test backend code, or frameworks, or proxies, or UIs—are beyond the scope of this document.
*I wrote this after not being able to find a similarly straightforward resource, that showed on what axes you could plot the different test types. I'd be happy to be corrected on anything substantial in this document, and even happier if some other document exists that supersedes it. PRs/issues/queries welcome.
