.NET NumericUpDown

Zaibot

It’s not that I hate Microsoft or anything... But I keep running in to strange things inside their code.

Before I continue, I just want to say to all of you how much fun Reflector can be ;-).

Possible thread sync errors, overflows and strange names these are the most common errors I run into in the .NET Framework.

This is just one example and thought I should post it.

Look at the code below. Nothing seems to be wrong right? This is an example of one of these strange overflows. When you enter decimal.MinValue and decimal.MaxValue into the minimum and maximum for a numeric control the overflow occurres, what happens is that it tries to calculate the amount of numbers required to display the string. Creating a decimal that contains 28 numbers that represent highest single digit (in this case 28 9’s) which results in an overflow.

public class NumericUpDown : UpDownBase, ISupportInitialize

{

/* ... */

internal override Size GetPreferredSizeCore(Size proposedConstraints)

{

decimal num5;

int preferredHeight = base.PreferredHeight;

int end = this.Hexadecimal ? 0x10 : 10;

int largestDigit = this.GetLargestDigit(0, end);

int num4 = (int)Math.Floor(Math.Log(Math.Max(-((double)this.Minimum), (double)this.Maximum), (double)end));

if ((largestDigit != 0) || (num4 == 1))

{

num5 = largestDigit;

}

else

{

num5 = this.GetLargestDigit(1, end);

}

for (int i = 0; i < num4; i++)

{

num5 = (num5 * end) + largestDigit;

}

int width = TextRenderer.MeasureText(this.GetNumberText(num5), this.Font).Width;

int num8 = base.SizeFromClientSize(width, preferredHeight).Width + base.upDownButtons.Width;

return (new Size(num8, preferredHeight) + this.Padding.Size);

}

/* ... */

}

Run this code below to see what happens:

using System;

using System.Windows.Forms;

public class WtfOverflow : Form

{

public WtfOverflow()

{

decimal num5;

int end = 10; //this.Hexadecimal ? 0x10 : 10;

int largestDigit = this.GetLargestDigit(0, end);

int num4 = (int)Math.Floor(Math.Log(Math.Max(-((double)decimal.MinValue), (double)decimal.MaxValue), (double)end));

if ((largestDigit != 0) || (num4 == 1))

{

num5 = largestDigit;

}

else

{

num5 = this.GetLargestDigit(1, end);

}

for (int i = 0; i < num4; i++)

{

num5 = (num5 * end) + largestDigit; // Overflow exception.

}

private int GetLargestDigit(int start, int end)

{

int num = -1;

int width = -1;

for (int i = start; i < end; i++)

{

char ch;

if (i < 10)

{

ch = i.ToString(CultureInfo.InvariantCulture)[0];

}

else

{

ch = (char) (0x41 + (i - 10));

}

Size size = TextRenderer.MeasureText(ch.ToString(), this.Font);

if (size.Width >= width)

{

width = size.Width;

num = i;

}

return num;

}

}

Ilya_Ehrenburg

I don't think it's a "strange" overflow, looks perfectly reasonable to me. I don't know .NET, so I may have the details wrong, but if I understand correctly, the maximal value for a base-10 decimal is 10^28-1, the minimal value is probably -(10^28-1) or -10^28. Converting to double gives ±1e28, abs, take the logarithm to base 10, get num4 = 28.

Now we have num5 == largestDigit = 9 = 10^1-1.

for (int i = 0; i < num4; i++)

 {

 num5 = (num5 * end) + largestDigit;

 }

i = 0: num5 = 10*9+9 == 10^2-1
i = 1: num5 = 10*(10^2-1)+9 == 10^3-1
etc. until i = 26: num5 = 10*(10^27-1)+9 == 10^28-1
i = 27: num5 = 10*(10^28-1)+9 == oops, overflow.

Be careful with doubles.

Zaibot

Yeah, it has just a tiny overflow. But the exception occurres, if they just used a double for the num5 variables there would have been no problem. Ahwell, just to make it easy I devided the decimals min and max by 10. I don't think we will use much larger then that =P, and it works now. Also noticed in their calculations they forget that there could be a negative symbol in the front.

Greetz

Weng

Could it perhaps be because that control is bloody fucking useless?