[Courses] [C] Floating Point Errors: summary
Mary
mary-linuxchix at puzzling.org
Thu Jul 18 11:25:31 EST 2002
I thought Suzi's post and the resulting discussion deserved its own
summary:
http://www.linuxchix.org/content/courses/c_programming/floating.html
Floating Point Errors
This informal discussion sprung from Suzi's [1]post describing a
problem with this program [code reproduced with permission]:
#include
char line[50];
float amount; /* the amount of $ as entered by the user */
int total; /* amount * 100, converted to an interger */
int dollars; /* # of dollars needed to make change */
int quart; /* # of quarters needed */
int dime; /* # of dimes needed, etc. */
int nick;
int penn;
int main()
{
while (1) {
/* inputting dollar amount */
printf("Enter the amount of money to make change for: $");
fgets(line, sizeof(line), stdin);
if (sscanf(line, "%f", &amount) <= 0)
break;
while (amount <= 0) {
printf("Please enter an amount greater than 0: $");
fgets(line, sizeof(line), stdin);
if (sscanf(line, "%f", &amount) <= 0)
break;
}
total = amount * 100;
/* determining how many of each coin is needed */
dollars = total / 100;
total %= 100;
quart = total / 25;
total %= 25;
dime = total / 10;
total %= 10;
nick = total / 5;
penn = total % 5;
/* printing the change needed. Program should only print a type of
* coin if it is needed to make change (i.e. it should not print
* "0 quarters". */
printf("To make change for $%.2f, you need:\n", amount);
if (dollars > 1)
printf("%d dollars\n", dollars);
if (dollars == 1)
printf("%d dollar\n", dollars);
if (quart > 1)
printf("%d quarters\n", quart);
if (quart == 1)
printf("%d quarter\n", quart);
if (dime > 1)
printf("%d dimes\n", dime);
if (dime == 1)
printf("%d dime\n", dime);
if (nick == 1)
printf("%d nickel\n", nick);
if (penn > 1)
printf("%d pennies\n", penn);
if (penn == 1)
printf("%d penny\n", penn);
}
return (0);
}
It didn't behave correctly when there was change of a single peeny (1
cent, or 0.01 dollars) involved.
Mary made a [2]long post in response which eventually identified a
floating point error, a quite subtle bug.
Online references
* [3]A general introduction
* [4]The source of this particular bug (the 0.1 bug)
Solutions
Different solutions to Suzi's problem were suggested:
* [5]Suzi's soluion [manually rounding to correct for floating point
errors]
* [6]Christopher's solution [using round or lround]
* [7]Conor's solution [correcting manual rounds for negative
numbers]
* [8]Lorne's solution [cutting floating point numbers out of the
equation altogether]
References
1. http://mailman.linuxchix.org/pipermail/courses/2002-July/000571.html
2. http://mailman.linuxchix.org/pipermail/courses/2002-July/000572.html
3. http://www.ug.cs.usyd.edu.au/~cs2/soft2004/s1_2002/tutes/week10/
4. http://www.eskimo.com/~scs/C-faq/q14.1.html
5. http://mailman.linuxchix.org/pipermail/courses/2002-July/000575.html
6. http://mailman.linuxchix.org/pipermail/courses/2002-July/000576.html
7. http://mailman.linuxchix.org/pipermail/courses/2002-July/000580.html
8. http://mailman.linuxchix.org/pipermail/courses/2002-July/000581.html
More information about the Courses
mailing list