Round Values

[mickey_pt]

Hello
I have an application that need to round some values, with a normal symetric round, but the options available in the round function doesn't allow me to do this kind of round.

For example, if i have

?Math.Round(0.275,2)
0.28
?Math.Round(1.275,2)
1.27
?Math.Round(1.275,2,MidpointRounding.AwayFromZero)
1.27
?Math.Round(1.275,2,MidpointRounding.ToEven)
1.27

The only way that looks like that i have a symmetric round it's when i use the format function:

?Format(0.275,"N2")
"0.28"
?Format(1.275,"N2")
"1.28"

I read some info available at msdn, i understand the bankers round technique, but when using the midpoint rounding the round should be ok, and it isn't. I'm missing something here?

Can i use the Format function and by using the Nx string, i have a symmetric rounding always?

Thanks

Meanwhile i have this function to do the job:

Public Function MyRound(ByVal valor As Decimal, ByVal casasDecimais As Integer) As Decimal
        Dim sinal As Integer = Math.Sign(valor)
        Dim escala As Decimal = Math.Pow(10, casasDecimais)
        Dim arredondado As Decimal = Math.Floor(Math.Abs(valor) * escala + 0.5D)
        Return ((sinal * arredondado) / escala)
    End Function

[Inferrd]

Hi,

The Math.Round method works on Decimal type values, but you are passing it Doubles in your examples

?Math.Round(CDec(0.275), 2)
'0.28
?Math.Round(CDec(1.275), 2)
'1.28
?Math.Round(CDec(1.275), 2, MidpointRounding.AwayFromZero)
'1.28
?Math.Round(CDec(1.275), 2, MidpointRounding.ToEven)
'1.28

Turning Option Strict On is a good idea. It will show some type conversion errors in your own round function, but unfortunately won't pick up the problem when you pass in literal values.

EDIT
Strike that. It of course takes Doubles, so I have no idea why it is rounding like it does in your examples.

[Inferrd]

A little more digging reveals that the problem is the precision with which floating point numbers are stored in Double type variables.

Try:

MessageBox.Show("1.275 is stored as " & 1.275.ToString("G17"))

More info here

[wes4dbt]

Micky_pt,


did you find which rounding method will produce the correct answer consistantly? I never knew Math.Round had that problem.

[i00]

I always did this:

Format(2.65, "0.#")

And for your function why wouldn't you use ....?:

Public Function MyRound(ByVal valor As Decimal, ByVal casasDecimais As Integer) As Decimal
    retrun Format(valor, "N" & casasDecimais.ToString))
End Function

Kris

[mickey_pt]

Thanks for all the answers.

@Inferrd
Thanks for looking for the answer for this "issue". In the end the answer it's that when the number isn't a binary fraction you may ou may not have the correct round.
But anyway the round function with the midpoint rounding should work, even if i use a double, and it doesn't, or I'm missing another "issue" here?!

@wes4dbt
Like i wrote, the method that gives me the correct value it's the Format, at least for the tests that i made, but i don't have 100% sure...

@i00
And that's why i wrote MyRound.

So after all the tests and all the reading, i came to the conclusion that i should use always decimal values when perform rounding, just to avoid some surprise in the end.

Thanks

[Jenner]

I found it best (somewhat the same way you did) that when dealing with low-precision decimal values, like currency or anything else up to around 5 decimal places max, using the "Decimal" type alleviated most of my headaches. It's precise and exact. Traditional float types like Single and Double work best when you need to grind a lot of numeric data fast, you're not doing anything like rounding, and you don't mind the value might be a little off.

For most business and industrial apps, I use Decimal exclusively. For things like games, Singles and Doubles work fine.

[dbasnett]

This article explains this behavior.

http://weblogs.asp.net/sfurman/archi...3/07/3537.aspx