When designing a method interface, a design recommendation that you often hear is to require as little as you have to to get the method done.
For example, Kent Beck wrote an article called "Test First isn't about testing" (http://computer.org/software/homepage/2001/05Design/index.htm) in which he initially had a MortalityTable class that requires a Person instance in its constructor. He later improved this to take an age as an integer, to decouple his MortalityTable class from the Person class.
This establishes the precendent of an age being an integer. Now, lets say later you require ages to be floats, to deal with some crazy sounding requirement involving ages like 48.2, for example. By tying ages to an integer representation, we are exposing ourselves to potentially large change propagation if the interpretation of age changes. This problem can be compounded if you go the extreme route of one requirement at a time.
So, the bottom line is:
* When to expose primitives and when to use Objects
Previous Topic
