Search

Saturday 7 May 2011

Rounding Floating-Point Numbers Up or Down in .NET


Introduction

Yesterday, I needed to be able not only to round a floating-point number to a certain number of decimal places, but also control the direction the rounding (i.e.: always force it to round down).

Not a problem, I thought. There will be a method on the System.Math class that will let me do just that. How wrong can you be?? My first port of call was the Math.Round() method, but whilst you can specify the number of decimal places with this method, you cannot force it to always round either up or down. Next, I looked at Math.Floor() and Math.Ceiling() methods. These, however, will only round to the nearest integer.

There was nothing else for it, but to roll my own.

The Solution

The code for my solution is as follows:

// C#
public static class EnhancedMath
{
    private delegate double RoundingFunction(double value);

    private enum RoundingDirection { Up, Down }

    public static double RoundUp(double value, int precision)
    {
        return Round(value, precision, RoundingDirection.Up);
    }

    public static double RoundDown(double value, int precision)
    {
        return Round(value, precision, RoundingDirection.Down);
    }

    private static double Round(double value, int precision, RoundingDirection roundingDirection)
    {
        RoundingFunction roundingFunction;
        if (roundingDirection == RoundingDirection.Up)
            roundingFunction = Math.Ceiling;
        else
            roundingFunction = Math.Floor;
        value *= Math.Pow(10, precision);
        value = roundingFunction(value);
        return value * Math.Pow(10, -1 * precision);
    }
}
' Visual Basic
Public Module EnhancedMath

    Private Delegate Function RoundingFunction(ByVal value As Double) As Double

    Private Enum RoundingDireciton
        Up
        Down
    End Enum

    Public Function RoundUp(ByVal value As Double, ByVal precision As Integer) As Double
        Return Round(value, precision, RoundingDireciton.Up)
    End Function

    Public Function RoundDown(ByVal value As Double, ByVal precision As Integer) As Double
        Return Round(value, precision, RoundingDireciton.Down)
    End Function

    Private Function Round(ByVal value As Double, ByVal precision As Integer, ByVal roundingDirection As RoundingDireciton) As Double
        Dim roundingFunction As RoundingFunction
        If roundingDirection = RoundingDireciton.Up Then
            roundingFunction = AddressOf Math.Ceiling
        Else
            roundingFunction = AddressOf Math.Floor
        End If
        value = value * Math.Pow(10, precision)
        value = roundingFunction(value)
        Return value * Math.Pow(10, -1 * precision)
    End Function

End Module

The rounding algorithm is quite simple: Firstly we shift all the digits we are interested in keeping, to the left of the decimal point. We then either call Math.Floor() or Math.Ceiling() on the result depending on whether we are rounding down or up respectively. Finally, we shift our digits back to the right of the decimal point and return the value.

The method handles values of type Double. You can always add your own overloads to handle values of other types(e.g.: Decimal).

Some Examples

Here are some examples of our code in action:

// C#
static void Main(string[] args)
{
    Console.WriteLine(EnhancedMath.RoundUp(3.141592654, 2)); // Result: 3.15
    Console.WriteLine(EnhancedMath.RoundUp(3.141592654, 3)); // Result: 3.142
    Console.WriteLine(EnhancedMath.RoundDown(3.141592654, 2)); // Result: 3.14
    Console.WriteLine(EnhancedMath.RoundDown(3.141592654, 3)); // Result: 3.141
    Console.Read();
}
' Visual Basic
Sub Main(ByVal args() As String)
    Console.WriteLine(EnhancedMath.RoundUp(3.141592654, 2)) ' Result: 3.15
    Console.WriteLine(EnhancedMath.RoundUp(3.141592654, 3)) ' Result: 3.142
    Console.WriteLine(EnhancedMath.RoundDown(3.141592654, 2)) ' Result: 3.14
    Console.WriteLine(EnhancedMath.RoundDown(3.141592654, 3)) ' Result: 3.141
    Console.Read()
End Sub

2 comments:

  1. Doens't work quite well.
    1.110m round up to 1.12m.

    ReplyDelete
  2. works for me on 0 precision. thanks

    ReplyDelete