Convert.ToInt32(double) strange rounding behaviour

  • Thread starter Thread starter Norvin Laudon
  • Start date Start date
N

Norvin Laudon

Hi,

Can somebody explain the following, from the MSDN documentation for the
"System.Convert.ToInt32(double)" function

<quote>
Return Value

value rounded to the nearest 32-bit signed integer. If value is halfway
between two whole numbers, the even number is returned; that is, 4.5 is
converted to 4, and 5.5 is converted to 6.
</quote>

Is there some obvious reason for this behaviour that I'm missing? Why would
one want 4.5 to be rounded to 4 rather than 5?

Confused,
Norvin
 
Norvin,

This is called "Banker's rounding". Googling for that should give you
more details.



Mattias
 
Norvin Laudon said:
Hi,

Can somebody explain the following, from the MSDN documentation for the
"System.Convert.ToInt32(double)" function

<quote>
Return Value

value rounded to the nearest 32-bit signed integer. If value is halfway
between two whole numbers, the even number is returned; that is, 4.5 is
converted to 4, and 5.5 is converted to 6.
</quote>

Is there some obvious reason for this behaviour that I'm missing? Why would
one want 4.5 to be rounded to 4 rather than 5?
Click to expand...

And, why would one want 4.5 to be rounded to 5 rather than 4?

From a mathematical standpoint, 4.5 is halfway. So why would rounding it up
make sense and rounding it down not make sense?

Rounding always in the same direction increases the chances of getting
strong deviations when you add lots of numbers that have been rounded. The
banker's rounding rule tries to reduce this undesirable effect.

Bruno
 
Also, be ready for some surprises if you use float or double (see other
posts on this ng). If you want predictable results for your rounding
operations, you should use the decimal type.

If you want to round half-up (or rather towards infinity, which means
rounding half-down on negative numbers), you can use the following:

static decimal RoundHalfTowardsInfinity(decimal dec) { return dec > 0 ?
(int)(dec + 0.5m) : (int)(dec - 0.5m); }

Side note for MS: it would be nice if the .NET framework provided rounding
methods for the various rounding modes, aka. Java's BigDecimal.

Bruno.
 
Back
Top