Lessons

Let's think about what lessons we can draw from this experiment.

The ability to use a function without having to know how it is implemented (and conversely, the ability to write a function without having to know how it will be used) is what makes writing large programs possible. There are a number of reasons for this.

• It makes it possible for more than one programmer to work on a single program at the same time. The programmers have to decide what functions are needed and agree on their interfaces. Then the functions can be implemented separately and combined to obtain a working program.

• It makes it possible to implement and test the pieces of a program individually. Suppose that two programs are of the same length, but one is broken up into functions and the other is not. It is much easier to get the first program working properly, because the small functions can be dealt with one at a time.

• It makes it possible to share code with other programmers. There are a number of national libraries of Fortran functions that solve common numerical problems. A programmer can take functions from these libraries and incorporate them into his or her programs. This is clearly superior to rewriting the functions from scratch.

You should adopt this approach in your own programs. When you approach a problem, first determine how it can be broken down into a main program and a collection of functions (and, as we will see in the next lesson, subroutines). Decide on the interface of each piece. Then work on one piece at a time, getting it to work properly (with stubs if necessary). Finally, put the pieces together and iron out any final problems.

There is one final thing that I'd like you to notice. In the original interface I said that the parameters to SOLVE were P and Q. In the final version of SOLVE that I gave you, notice that the parameter names are A and B. The exact names of the parameters of a function do not matter when you are calling a function. Be sure that you understand why.