Unit tests should test all usual and also all special cases of the functionality. If there is a special case of code that is not covered by the unit test, the code is in danger. In case of a sort implementation, the general case is sorting, say five elements. The special cases are much more numerous usually. How does our code behave if there is only one element or if there are no elements? What if there are two? What if the elements are in reverse order? What if they are already sorted?

Usually, the special cases are not defined in the specification. The programmer has to think about it before coding, and some special cases are discovered during coding. The hard thing is that you just cannot tell if you covered all special cases and the functionality of the code.

What you can tell is if all the lines of code were executed during the testing or not. If 90% of the code lines are executed during the tests, then the code coverage is 90%, which is fairly good in real life, but you should never be content with anything less than 100%.

Code coverage is not the same as functional coverage, but there is a correlation. If the code coverage is less than 100%, then at least one of the following two statements is true:

• The functional coverage is not 100%
• There is unused code in the tested unit, which can just be deleted

The code coverage can be measured, the functional coverage cannot. The tools and IDEs support code coverage measurement. These measurements are integrated into the editor so you will not only get the percentage of the coverage, but the editor will show you exactly which lines are not covered by the coverage coloring the lines (in Eclipse, for example) or the gutter on the left side of the editor window (IntelliJ). The picture shows that in IntelliJ, the tests cover the lines indicated by a green color on the gutter. (In the print version this is just a grey rectangle).