The idea behind Log Driven Testing is to make some assertions on the log messages that should be output by the code you test.
Before reading further, it is important to understand that this idea originally comes from an observation: log messages are often missing, or they don't provide enough information. Sometimes, they are even wrong and can misslead investigations in case of problems. The aim of Log Driven Testing is to fix this, while at the same time providing a new way of writting automated tests.
Usually, when we write automated tests, we make assertions on the results of a function (what it returns), or on the state of some objects that may be affected by the function, or even on calls to other functions that may be called by the function we are testing.
Log Driven Testing has a different approach, in that it doesn't assert this usual stuff, but instead asserts logs. The purpose is not necesseraly to replace classical tests, but it can advantageously complement them.
Main advantages could be:
- You are obliged to write proper log messages to have good tests.
- Log messages are also tested.
- Assertions are easy to read as they are plain text (log messages).
- Assertions are very easy to write.
- Multiple level of testing can be achieved thanks to different log levels.
Log Driven Testing must be tried in a real-world application in order to see how it behaves and if it brings any productivity (and/or robustness) gain.
For this purpose, I am implementing a simple Go library to make it simple to write such tests.