As long as you are testing functions that do not interact too much with the outside world, things are pretty easy. The problems start when the units you are testing interface with third-party components such as databases, HTTP connections, and specific files.
On one hand, you want to see how your code behaves due to various circumstances. On the other hand, you don't want to wait for the database to boot, and you definitely don't want to have several databases containing different versions of data so that you can check all the necessary conditions.
How can we deal with such cases? The idea is not to execute the actual code that triggers all those side effects but instead use test doubles. Test doubles are constructions in code that mimic the actual API, except they don't perform actions of the mimicked functions or objects.
The most common test doubles are mocks, fakes, and stubs. Many people tend to mistake one for another as they are similar...
