Floating Point Number errors can sink your app with issues - Essential Software Development Career + Technical Guide

Floating point errors cause numbers not to behave as you would expect.

You would think that 1/7 added seven times would equal 1. In normal math, it would, but with double floating point math, computers use no.

Let’s take this example below comparison. Do you think this should be equal?

(1.0/7.0) + (1.0/7.0) + (1.0/7.0) + (1.0/7.0) + (1.0/7.0) + (1.0/7.0) + (1.0/7.0) == 1.0

The math you learned in grade school unfortunately doesn’t hold true for floating point math. That comparison done with double variables in C# would not be equal.

That’s because adding 1/7 together seven times equals 0.99999999999999978.

With floating point math, there is the opportunity to accumulate errors in the least significant bits of the double type.

The best thing to do in these instances is to decide the precision you need and truncate to within the supported precision and round if desired.

For example review the below code that rounds to a specific prevision:

if (Math.Round(((1.0 / 7.0) + (1.0 / 7.0) + (1.0 / 7.0) + (1.0 / 7.0) + (1.0 / 7.0) + (1.0 / 7.0) + (1.0 / 7.0)), 15)==1.0)
                Console.WriteLine(“Equal”);  //would be equal
            else
                Console.WriteLine(“Not Equal”);

This ensures we would get the expected answer of equal in this case, as it will round up to match the expected value here.

An accepted answer on StackOverflow for comparing Equal:

bool CloseToEqual(double a, double b) {
    double epsilon1 = Math.Max(Math.Abs(a), Math.Abs(b)) * 1E-15;
    return Math.Abs(a - b) <= epsilon1;
}

Note: Method/variable names were changed. See “C#—Double.Epsilon for Equality, Greater than, Less than, Less than or Equal to, Greater than or Equal to,” Stack Overflow, n.d.)

A variation of this article appeared in the book Essential Software Development Career + Technical Guide.

If you would like to understand the details of why read this detailed article Floating Point in .NET part 1: Concepts and Formats.

If you need precise calculations for things like Money/Financial transactions, consider the C# Decimal datatype as an option. Depending on your usage requirements/performance, one may make more sense or the other.

See References for more information relating to floating point errors:

“C#—Double.Epsilon for Equality, Greater than, Less than, Less than or Equal to, Greater than or Equal to”—Stack Overflow. (n.d.). Retrieved September 17, 2022, from https://stackoverflow.com/questions/2411392/double-epsilon-for-equality-greater-than-less-than-less-than-or-equal-to-gre. License:https://creativecommons.org/licenses/by-sa/2.5/ and https://creativecommons.org/licenses/by-sa/3.0/)

Visit our home page here.

Thanks for reading. Sign up for our email list on the right to get notified about the articles we post. Check out our other blog articles.

Related posts: