PersonNinja PersonNinja - 3 months ago 15
C Question

Why is this change counting algorithm acting unreliably?

I'm somewhat new to programming, only having taken one course on Python at university and now working my way through Harvard's CS50 OpenCourseware, so please bear with me.

This code compiles fine, no errors, etc. The program is meant to take a user input amount of change and use a simple greedy algorithm to return the fewest number of each U.S. coin it would take to represent that change. Simple enough; however, what's killing me here is that, for some reason, it doesn't always count the pennies.

If I were to enter ".41", I would get "1 Quarters, 1 Dimes, 1 Nickels, and 0 Pennies"; entering
".42" yields "1 Quarters, 1 Dimes, 1 Nickels, and 1 Pennies";
but oddly ".43" yields the correct "1 Quarters, 1 Dimes, 1 Nickels, and 3 Pennies".

It's the inconcistency of this bug that's making it so hard for me to track down. I keep on walking through the code in my head with different inputs to try to pinpoint the problem but it's been a futile effort.

What am I doing wrong?

#include <stdio.h>
#include <cs50.h>

int main(void)
{
printf("How much change is owed? ");
float change = GetFloat();
int quarters = 0;
int dimes = 0;
int nickels = 0;
int pennies = 0;

float coinArray[4] = {.25, .10, .05, .01};
int coinNames[4] = {quarters, dimes, nickels, pennies};

int counter(float coinArray);
{
int x;
for(x = 0; x < 4; x++)
{
while (change >= coinArray[x])
{
change = change - coinArray[x];
coinNames[x]++;
}
}

}
printf("%d Quarters, %d Dimes, %d Nickels, and %d Pennies\n",
coinNames[0], coinNames[1], coinNames[2], coinNames[3]);
}

Answer

It's the simple fact that your computer simply cannot represent 0.2 accurately. Mathematically it's just 1/5, but your computer can't represent that as a float in the same way that you can't represent 1/3 as a decimal number. 0.41 is just 41/100 which is a minimal fraction and has two factors of five in the denominator - no chance to represent as a float.

The only way to handle currency accurately, is to calculate in terms of cents, not dollars. That way you use integers (int, or better long long instead of float), and never get roundoff errors.

Comments