Comparing currency values as floats does not work

Question

If I type 0.01 or 0.02 the the program just exits immediately. What do I do?

#include <stdio.h>

char line[100];
float amount;

int main()
{
printf("Enter the amount of money: ");
fgets(line, sizeof(line), stdin);
sscanf(line, "%f", &amount);

if (amount == 0.0)
    printf("Quarters: 0\nDimes: 0\nNickels: 0\nPennies: 0");

if (amount == 0.01)
   printf("Quarters: 0\nDimes: 0\nNickels: 0\nPennies: 1");


if (amount == 0.02)
   printf("Quarters: 0\nDimes: 0\nNickels: 0\nPennies: 2");

return (0);
}

Answer 1

What is the most effective way for float and double comparison?

You should use something like this

bool isEqual(double a, double b)
{
    return fabs(a - b) < EPSILON;
}

or this

#define ISEQUAL(a,b) ( fabs((a) - (b)) < EPSILON)

and define EPSILON

#define EPSILON (0.000001)

Answer 2

Firstly, do not attempt to compare doubles, or to compare floats, or to compare doubles to floats (exception: comparing against 0 is OK, and arguably all integers). Other answers have given good links as to why, but I especially liked: http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html

Secondly, if you are dealing with an amount of money, don't use doubles, or get out of double as quickly as possible. Either use a fixed point library (a bit complex for what you want), or simply do:

int cents = (int)round(amount * 100);

then work with cents instead. You can compare these to your heart's content.

Thirdly, if the idea here is to work out change in quarters, dimes, nickels and pennies, a collection of if statements is not the way to go, unless you like long and inefficient programs. You will need to do that programatically. And that will be far easier if you have the integer number of cents above (hint: look at the % operator).

Answer 3

I think we could scale for a factor and make a relative comparison like this:

#define EPSILON (0.000001)

if (fabs(a - b) <= EPSILON * fmax(fabs(a), fabs(b))) {
}

which is more robust than an absolute comparison:

if (fabs(a - b) <= EPSILON ) {
    // ...
}

which maybe scale dependent. By the way having the optimum scale depends also from the function, we shouldn't use the same formula for a log and for a sin.

Answer 4

The downvotes on this answer are well deserved, it was wrongly advocating for a simple equality test. While this is undefined behaviour and should not be encouraged, even so it happens to work sometimes.

You should never test a floating value for equality. Instead use an epsilon value to test your variable against a range as this answer is suggesting it.

For instance take this example:

#include <stdio.h>
#include <math.h>

#define  EPSILON 0.00001f

int main() {
    float a = 3.234324333f;
    float b = a;
    b += 1;
    b -= 1;
    
    printf("a value is %f\n", a);
    printf("b value is %f\n", b);

    printf("\na == b ?\n");
    if (a == b) {
        printf("true\n");
    } else {
        printf("false\n");
    }

    printf("a ~= b with epsilon = %f ?\n", EPSILON);
    if (fabs(a - b) <= EPSILON) {
        printf("true\n");
    } else {
        printf("false\n");
    }
    return 0;
}

When running it:

a value is 3.234324
b value is 3.234324

a == b ?
false
a ~= b with epsilon = 0.000010 ?
true