Question

My boss reported my a bug today because my configuration was decreasing whenever he wanted to specify a value below 5%. I know that I can just round my number before casting it as int to fix my problem, but I don’t understand why this problem occurs.

I have a app.config file with the value “0.04” and a configuration section with a float property. When the section is read, the float value retrieved is 0.04, which is fine. I want to put this value in a windows forms TrackBar which accept an integer value so I multiply my value by 100 and a cast it as int. For some reason, the result is not 4, but it’s 3. You can test it like this :

Console.WriteLine((int)(float.Parse("0.04", System.Globalization.CultureInfo.InvariantCulture) * 100)); // 3

What happened?

Answer 1

It’s because 0.04 can’t be exactly represented as a float – and neither can the result of multiplying it by 100. The result is very slightly less than 4, so the cast to int truncates it.

Basically, if you want to use numbers represented accurately in decimal, you should use the decimal type instead of float or double. See my articles on decimal floating point and binary floating point for more information.

EDIT: There’s something more interesting going on here, actually… in particular, if you assign the result to a local variable first, that changes the result:

using System;
using System.Globalization;

class Test
{
    static void Main()
    {
        // Assign first, then multiply and assign back, then print
        float f = Foo();
        f *= 100;
        Console.WriteLine((int) f); // Prints 4

        // Assign once, then multiply within the expression...
        f = Foo();
        Console.WriteLine((int) (f * 100)); // Prints 4

        Console.WriteLine((int) (Foo() * 100)); // Prints 3
    }

    // No need to do parsing here. We just need to get the results from a method
    static float Foo()
    {
        return 0.04f;
    }
}

I’m not sure exactly what’s going on here, but the exact value of 0.04f is:

0.039999999105930328369140625

… so it does make sense for it not to print 4, potentially.

I can force the result of 3 if the multiplication by 100 is performed with double arithmetic instead of float:

f = Foo();
Console.WriteLine((int) ((double)f * 100)); // Prints 3

… but it’s not clear to me why that’s happening in the original version, given that float.Parse returns float, not double. At a guess, the result remains in registers and the subsequent multiplication is performed using double arithmetic (which is valid according to the spec) but it’s certainly a surprising difference.