Stroutrup's 20-year-old advice even more sound today . . .

"C++ as a Better C" revisited

© Conrad Weisert, Information Disciplines, Inc., June 25, 2009

NOTE: This document may be circulated or quoted from freely, as long as the copyright credit is included.
.

Why write a procedural program in C++?

In introducing C++ Bjarne Stroustrup pointed out that some of the new features would be useful even to programmers who didn't exploit any object-oriented techniques.

At that time, we could still buy a pure C compiler from a mainstream vendor. It cost less and would run on a much smaller machine than the new C++ monsters. A programmer might well prefer some new C++ feature, but had to forgo it to conform to budget or configuration constraints.

All that has changed today. We all have access to affordable C++ compilers and sophisticated development platforms. We all have generous machine configurations that easily run them. So, with all that object-oriented power available, why would anyone want to write a program using only the procedural paradigm?

Integrating code with a legacy application, stuffing it into a tiny embedded processor, or complying with some strange contract are possible, but unlikely reasons. I have a more persuasive and more common reason: to teach introductory programming.

We want the beginning programmer to master procedural programming, but we don't want to burden him or her with unnecessary rules and restrictions. So, the beginning student will use a C++ compiler, but will ignore both

What oddities?

Here are some elements of C1 that I wouldn't mention to the student, along with corresponding features of C++ that require no OOP awareness:

Ugly error-prone C facility

Better C++ facility

malloc, free, and sizeof functions that require otherwise unnecessary knowledge of internal data representations and invite cheating. new and delete operators.
Passing a pointer to allow a function to modify its argument. Use reference parameter.
Manipulation of null-terminated string (array of char) either directly or through library routines strcpy and strcat See Coping with Character Strings in C (not a C++ facility, just good practice.2)
printf and fprintf for stream output, with associated format codes. Output-stream insertion operator <<.3
Similarly, scanf and fscanf for stream input, except where an extremely unusual format must be handled. Input-stream extraction operator >>3.
Preprocessor (macro) definition of a pseudo-generic function Define an inline function or function template.
Repeating the word struct when delaring an instance. Just use the name of the struct.2
Unique names for each type of non-macro generic function Function name overloading

With those simplifications, C++ turns out to be a fine language for procedural programming, and is well suited to teaching programming to beginners. Most compilers give excellent error diagnostics, and with suitable care run-time loss of control should be rare.


1—The C language continues to grow and change. The details mentioned here apply to the language as defined in Harbison & Steele: C—a Reference Manual, fourth edition, Prentice Hall, ISBN 0-13-326224-3.

2—These techniques will also help prepare the student for OOP.

3—Yes, these IO stream facilities depend on classes and objects, but the programmer needn't be aware of them. We can simply present the syntax as another kind of expression statement.

Return to IDI home page
technical articles
C/C++ articles

Last modified July 9, 2009