4.6. General function writing guidelines

1. Write for clarity and correctness first

2. Avoid premature optimization

3. Avoid premature “pessimization” That is, prefer faster when equally clear

4. Minimize side-effects

   A function that modifies its parameters is said to have side-effects. Programs with too many side-effects are hard to predict and debug.

   Returning to our call-stack example. What if the function signatures were changed to accept a pass-by-reference parameter?

   void dig(double& x);
   void deeper(double& x);
   

   Given that the names of these functions provide no insight to their purpose, there is no way to know without inspecting the source if the variable x is modified when passed to these functions.

   void dig(double& val) {
     std::cout << "Digging...\n";
     val /= 2;
     deeper(val);
     std::cout << "Done digging...\n";
   }
   

   Unless it’s obvious from the name of the function, side effects are almost always unexpected.

   #include <iostream> void dig(double& val); void deeper(double& val); int main() { double pi = 3.14159; std::cout << "in main.\npi = " << pi << '\n'; dig(pi); std::cout << "Returned to main.\npi = " << pi << '\n'; return 0; } void dig(double& val) { std::cout << "Digging...\n"; val = val * 2; deeper(val); std::cout << "Done digging...\n"; } void deeper(double& val) { val -= 1; std::cout << "now even deeper....\n"; }

   This is one of the reasons why some programmers only use pass-by-reference when the parameter is const. Some programmers prefer passing pointers over non-const parameters. This requires the caller to explicitly pass in an address and clearly states that the function may modify the parameter.

5. Keep functions short

   • A function should do one thing well

     If you see a function doing more than one thing consider breaking it up into multiple functions

   • Is this (slightly) more work?

     In the short run, perhaps.

     In the long run, your total time spent debugging, testing, maintaining, and modifying will be far, far less than if you packed everything into one monster function

   • Note that unit testing is practically impossible once functions reach a certain size.

6. Strive to write a function once and never modify it again.

7. Check function parameters for validity. Unless you completely trust the caller (and their caller…)

   • It should be obvious: do not trust argv[]

4.6.1. When to write a function

As with any kind of abstraction, there are two goals to making a function:

• Encapsulation: If you have some task to carry out that is simple to describe from the outside, but messy to understand from the inside, then wrapping it in a function lets the caller carry out this task without having to know the details.

  This is also useful if you want to change the implementation later.

• Code re-use: If you find yourself writing the same lines of code in several places (or worse, are tempted to copy a block of code to several places), you should probably put this code in a function (or perhaps more than one function, if there is no succinct way to describe what this block of code is doing).

Both of these goals may be trumped by the goal of making your code clear. If you can’t describe what a function is doing in a single, simple sentence, this is a sign that maybe you need to restructure your code. Having a function that does more than one thing (or does different things depending on its arguments) is likely to lead to confusion.

So, for example, this is not a good function definition:

// This code is an anti-pattern.
// It's an example of how NOT to write a function.

/**
 * If getMax is true, return maximum of x and y,
 * else return minimum.
 */
int computeMinOrMax(int x, int y, bool getMax) {
  if(x > y) {
    if(getMax) {
      return x;
    } else {
      return y;
    }
  } else {
    if(getMax) {
      return y;
    } else {
       return x;
    }
  }
}

This function is clearly trying to do two things and not doing either one very well. Two functions would be far simpler:

// return the maximum of x and y
// if x == y, return y
int maximum (int x, int y) {
  return (x < y) ? y : x;
}

// return the minimum of x and y
// if x == y, return y
int minimum (int x, int y) {
  return (y < x) ? y : x;
}

int computeMinOrMax(int x, int y, bool getMax) {
    if(getMax) {
      return maximum(x, y);
    }
    return minimum(x, y);
}

// or more compactly:
int computeMinOrMax(int x, int y, bool getMax) {
  return getMax ? maximum(x,y) : minimum(x,y);
}

Is this slightly more typing? Yes. At the end of the day, you will be far happier testing and debugging the three simpler functions than the first version. Your future co-workers will thank you.

Note

Also be aware the STL provides functions std::min and std::max, which eliminate the need for our minimum and maximum entirely and have the advantage of working on any comparable type.

This would transform the previous example to:

#include <algorithm>

int computeMinOrMax(int x, int y, bool getMax) {
  return getMax ? std::max(x,y) : std::min(x,y);
}

4.6.2. Example: number guessing

A more realistic example might help.

While randomly surfing the internet I stumbled on a small program intended to help people understand C++.

Unfortunately, it is full of issues and this is the kind of program structure that will not help you when trying to create your own complicated projects.

Open this in Replit and run ./main in the console tab. You may need to type make main first.

Try This!

How many bugs or other issues can you find in this program without looking at the ‘Bugs’ tab?

This program has a few bugs.

Code like:

if(tries >= 20)

seems to imply you have 20 tries. The program actually gives you 21. Off-by-one errors like this are common.

This code looks ok, but isn’t.

std::cout << "Enter a number between 1 and 100 (" << 20 - tries << " tries left): ";
std::cin >> guess;
std::cin.ignore();

There is no error checking on guess before it is used. Non-integer input causes the program to enter an infinite loop.

Related to this, there is this code:

int guess;

// try to get guess

if(guess > number)

Since guess is uninitialized, if cin fails to fill guess, then guess will not have any value when the if statement is evaluated, which is undefined behavior.

This is fundamentally a C program on a C++ forum.

Yes, it uses cin and cout.

That doesn’t make it a C++ program.

A clue it was copied from C:

int main (void) . . .

Explicitly using void to declare a function take no parameters is a best practice in C. Otherwise the compiler assumes the function can take any number of parameters.

In C++, main() or any other function that takes no parameters is implicitly void.

Next problem is the way the random numbers are created:

int number = rand() % 99 + 2;

Quick!

Can we be certain that this correctly creates a number from 1 to 100, inclusive?

The code is just not that easy to reason about.

• We have to know how rand works

• We have to remember what modulus does.

Yes, not big hurdles, but this is where bugs hide. And for the record, the program asks the user to pick a number from 1 to 100, but this algorithm will never choose 1.

The standard library has a superior alternative to rand:

#include <iostream> #include <random> int main() { std::cout << "Random numbers: \n"; // Seed with a real random value, if available std::random_device r; std::default_random_engine eng(r()); // make a random number generator for (int i = 0; i < 10; ++i) { // generate the next uniformly distributed integer between 0 and 999 // using the random default engine std::cout << std::uniform_int_distribution<int> {0, 999} (eng) << '\n'; } }

Because there are no functions, it is necessary to repeat block of code like this:

std::cout << "Enter a number between 1 and 100 (" << 20 - tries << " tries left): ";
std::cin >> guess;
std::cin.ignore();

In general, the pattern

• Prompt

• Assign

• Validate

• Repeat (if needed) or exit

Is common. Because it’s tedious to copy over and over, this program omits the error handling and the repeat.

A function is the obvious choice here.

Try This!

Run the program on the original tab, but enter a letter or other nonsense input instead of a number.

How would you fix this? Try it!

More repetition. How many times is the number 20 used in this program?

If you wanted to change the max number of guesses to 10, how many places do you need to remember? And this is just one file. These kinds of duplications can become painful to maintain as programs grow. They can quickly get out of control.

Finally, pet peeves of mine:

This code is mostly redundant. I just prefer not to see code that looks like this, even though it works.

if(answer == 'n' || answer == 'N' || answer == 'y' || answer == 'Y') {

I would rather use a function:

char play_again() {
  return std::tolower(
      get_entry("Would you like to play another game? [y/n] ").front());
}

// and use it like this
while ('y' == play_again());

instead of:

while(true) { // Loop to ask user is he/she would like to play again.
   // Get user response.
   std::cout << "Would you like to play again (Y/N)? ";
   std::cin >> answer;
   std::cin.ignore();

   // Check if proper response.
   if(answer == 'n' || answer == 'N' || answer == 'y' || answer == 'Y') {
      break;
   } else {
      std::cout << "Please enter \'Y\' or \'N\'...\n";
   }
}

Is this comment helping?

// Get number.
               std::cout << "Enter a number between 1 and 100 (" << 20 - tries << " tries left): ";

// Safely exit.
std::cout << "\n\nEnter anything to exit. . . ";
std::cin.ignore();

Please don’t ask me to enter an additional confirmation to exit, when I just said ‘No’ to the previous question.

There is no need to do this. What is safe about this? Just exit your program.

Open this in Replit and run ./main in the console tab. You may need to type make main first.

This is NOT the only way to improve the original program. It’s merely one way.

Notice the finished program isn’t shorter than the original. That was not our goal. It rarely is. We made the program clearer and easier to reason about. While we were at it, we fixed some bugs and made it a bit more reusable and maintainable.

---

More to Explore

• C++ Core Guidelines for functions

• Unit testing library list

• A very brief description of “extract method” from Martin Fowler’s Refactoring site.

• ExtractMethod discussion from the PortlandPatternRepository - the very first wiki

You have attempted of activities on this page