Double minus zero

[Avin Patel]

Hi,
Does the Double has the facilty to define -0.

If it has, "==" or System.Math.Sig n() doesn't able to differentiate between -0 & 0.

i.e.
Double x = -0.0d
Double y = 0.0d

if(value == -0.0d) {//both x & y drops in this statement.
//is minus zero
}
else {
//Not minus zero
}

How to differentiate -0.0d & 0.0d in if statement?

Thank you,
Avin Patel

[Frank Oquendo]

Avin Patel wrote:

How to differentiate -0.0d & 0.0d in if statement?

How can a sign have any meaning when discussing the absence of a value?

--
Things never go according to plan. So plan accordingly.

[mikeb]

Avin Patel wrote:

Hi,
Does the Double has the facilty to define -0.

If it has, "==" or System.Math.Sig n() doesn't able to differentiate between -0 & 0.

i.e.
Double x = -0.0d
Double y = 0.0d

if(value == -0.0d) {//both x & y drops in this statement.
//is minus zero
}
else {
//Not minus zero
}

How to differentiate -0.0d & 0.0d in if statement?

Yup - Math.Sign() is documented to return 0 if the double value is 0
(and positive zero is equal to negative zero according to the IEEE spec).

To determine if you have negative zero, I think you'll have to write a
bit of code that looks at the sign bit directly:

public static bool IsSignBitSet( double x) {
byte[] bytes = BitConverter.Ge tBytes( x);

// this assumes a little-endian machine
return ((bytes[bytes.Length-1] & 0x80) == 0x80);
}
// ...

Console.WriteLi ne( "IsSignBitS et( +0.0): {0}", IsSignBitSet( +0.0));
Console.WriteLi ne( "IsSignBitS et( -0.0): {0}", IsSignBitSet( -0.0));

The above code outputs:

IsSignBitSet( +0.0): False
IsSignBitSet( -0.0): True

--
mikeb

[Jon Skeet [C# MVP]]

mikeb <ma************ @nospam.mailnul l.com> wrote:

How to differentiate -0.0d & 0.0d in if statement?

Yup - Math.Sign() is documented to return 0 if the double value is 0
(and positive zero is equal to negative zero according to the IEEE spec).

To determine if you have negative zero, I think you'll have to write a
bit of code that looks at the sign bit directly:

Alternatively, check for whether it's equal to positive 0 but has a
different bit sequence to positive 0 (still using
BitConverter.Ge tBytes, but regardless of endianness).

--
Jon Skeet - <sk***@pobox.co m>
http://www.pobox.com/~skeet
If replying to the group, please do not mail me too

[Jon Skeet [C# MVP]]

Frank Oquendo <no******@here. com> wrote:

How to differentiate -0.0d & 0.0d in if statement?

How can a sign have any meaning when discussing the absence of a value?

Very easily. -0 and +0 are mathematically distinct when considering
limits, where something can tend to 0 from above or below, for
instance. It's all well-defined in IEEE arithmetic, I believe.

--
Jon Skeet - <sk***@pobox.co m>
http://www.pobox.com/~skeet
If replying to the group, please do not mail me too

[mikeb]

Jon Skeet [C# MVP] wrote:

mikeb <ma************ @nospam.mailnul l.com> wrote:

How to differentiate -0.0d & 0.0d in if statement?

Yup - Math.Sign() is documented to return 0 if the double value is 0
(and positive zero is equal to negative zero according to the IEEE spec).

To determine if you have negative zero, I think you'll have to write a
bit of code that looks at the sign bit directly:

Alternatively, check for whether it's equal to positive 0 but has a
different bit sequence to positive 0 (still using
BitConverter.Ge tBytes, but regardless of endianness).

Alright - in the interest of endian-safeness, here's an alternative:

private readonly static long signBit =
BitConverter.Do ubleToInt64Bits ( -0.0);

public static bool IsSignBitSet( double x) {
long bits = BitConverter.Do ubleToInt64Bits ( x);

return( (bits & signBit) != 0);
}
--
mikeb

[Avin Patel]

Hi Mikeb,
But compact C# doesn't have " BitConverter.Do ubleToInt64Bits ()" functionality?

"mikeb" wrote:

Jon Skeet [C# MVP] wrote:

mikeb <ma************ @nospam.mailnul l.com> wrote:

How to differentiate -0.0d & 0.0d in if statement?

Yup - Math.Sign() is documented to return 0 if the double value is 0
(and positive zero is equal to negative zero according to the IEEE spec).

To determine if you have negative zero, I think you'll have to write a
bit of code that looks at the sign bit directly:

Alternatively, check for whether it's equal to positive 0 but has a
different bit sequence to positive 0 (still using
BitConverter.Ge tBytes, but regardless of endianness).

Alright - in the interest of endian-safeness, here's an alternative:

private readonly static long signBit =
BitConverter.Do ubleToInt64Bits ( -0.0);

public static bool IsSignBitSet( double x) {
long bits = BitConverter.Do ubleToInt64Bits ( x);

return( (bits & signBit) != 0);
}
--
mikeb

[Jon Skeet [C# MVP]]

Avin Patel <Av*******@disc ussions.microso ft.com> wrote:

But compact C# doesn't have " BitConverter.Do ubleToInt64Bits ()" functionality?

In that case, you can use BitConverter.Ge tBytes and just compare the
contents of the returned array with BitConverter.Ge tBytes(0d).

--
Jon Skeet - <sk***@pobox.co m>
http://www.pobox.com/~skeet
If replying to the group, please do not mail me too

[Avin Patel]

Hi,
I guess this should be ok for compact case:
static public bool IsMinusZero(dou ble data)
{
byte[] minuszero =
BitConverter.Ge tBytes(-0.0);

byte[] bytes = BitConverter.Ge tBytes(data);

for(int i=0; i<minuszero.Len gth; i++) {
if(bytes[i] != minuszero[i]) { return false; }
}
return true;
}

Thanks for replies,
Avin Patel

"Avin Patel" wrote:

Hi Mikeb,
But compact C# doesn't have " BitConverter.Do ubleToInt64Bits ()" functionality?

"mikeb" wrote:

Jon Skeet [C# MVP] wrote:

mikeb <ma************ @nospam.mailnul l.com> wrote:

>>How to differentiate -0.0d & 0.0d in if statement?
>
>Yup - Math.Sign() is documented to return 0 if the double value is 0
>(and positive zero is equal to negative zero according to the IEEE spec).
>
>To determine if you have negative zero, I think you'll have to write a
>bit of code that looks at the sign bit directly:
Alternatively, check for whether it's equal to positive 0 but has a
different bit sequence to positive 0 (still using
BitConverter.Ge tBytes, but regardless of endianness).

Alright - in the interest of endian-safeness, here's an alternative:

private readonly static long signBit =
BitConverter.Do ubleToInt64Bits ( -0.0);

public static bool IsSignBitSet( double x) {
long bits = BitConverter.Do ubleToInt64Bits ( x);

return( (bits & signBit) != 0);
}
--
mikeb