Font Awesome Free 5.13.0 by @fontawesome - License - (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License)

Testing patterns

Some useful patterns for writing tests or making production code more testable

Humble Object pattern

  • Problem: some behavior is hard to test
    • Example: the way data is presented in the UI
  • Solution:
    • Extract as much of the behavior as possible into an easy-to-test part
    • What's left is a Humble Object containing the hard-to-test stuff stripped down to the bare essence

See also separate note for Humble Object pattern

Clock Wrapper pattern

  • Problem: Behavior depending on current time can be hard to test
  • Solution: Don't get current time directly, but access it through some kind of wrapper that allows you to manually set the time to a specific value for testing purposes
  • Alternative: Pass in current time as a parameter for code you want to keep easy to test (more functional approach)

Wrapping can also be used for other dependencies that are hard to control or mock

Object Mother pattern

  • Problem: creating data for tests can be tedious and complex
    • You may need to create a complex hierarchy of objects before you can create object you are interested in
  • Solution: set up an "Object Mother" that creates predefined standard objects with all required data
    • Can make sense to have a few standard objects of the same type to represent variations
      • Example: fresh hire "John" and long-time employee "Heather"
    • Standard objects become familiar to the team and sometimes even enter functional discussions
      • Similar to personas, but not always people: can also be insurance policies, supplier contracts, ...
    • Often, standard objects need a bit of additional setup for the specific test
      • Avoid adding new standard objects, focus on small set that team is familiar with
    • Can have different Object Mothers for different object types
  • Drawback: coupling
    • Many tests will depend on the exact data specified in the Object Mother
    • Changing the standard data can have large impact