ansaurus

Question

comparing two floats to see if they're both negative, or both positive.

Answer 1

+9 A:

Multiply them together.

If the answer is positive then they are both the same sign.

If the answer is negative then they are of opposite sign.

If the answer is zero (within some value to take care of rounding error) then one or both are zero and you'll have to check them individually. You'll then have to decide whether 0 to be treated as positive or negative in your scenario.

ChrisF 2010-01-06 14:51:17

what about zero?

Simon 2010-01-06 14:55:16

ChrisF 2010-01-06 14:56:03

looking at it again myself, the OPs rule is ambiguous since zero is neither +ve or -ve.

Simon 2010-01-06 14:58:52

if they're both zero, thats fine. Would this solution work?

dotty 2010-01-06 15:00:34

@dotty yes, it would work, and it is a neat solution except when one or other value is zero. What does your rule say when only one value is zero?

Simon 2010-01-06 15:04:43

@dotty - if they are both zero then the answer will be zero (actually within rounding error of zero, but that's another matter). In any case a zero result will mean you have to check each input individually to see if it's positive or negative.

ChrisF 2010-01-06 15:05:39

Not sure, would it pass if the values are 0 and -3, or 0 and +3?

dotty 2010-01-06 15:06:34

@dotty - that's up to you to decide - is 0 to be treated as positive or negative in your scenario?

ChrisF 2010-01-06 15:11:22

Well since we're talking about IEEE standard floating point we have +0 and -0 as two separate numbers... although I'm not sure how you'd detect them without doing type puning and comparing them as int's (this has a potential issue in that there are insane platforms that use a different byte order for ints and floats...)

Spudd86 2010-06-14 15:34:34

Answer 2

+2 A:

Just do something like:

if float1*float2<0
    Error

smoore 2010-01-06 14:51:41

Answer 3

A:

function isSameSign(a,b){
  var r = a*b; 
  return (r>0 || r==0)
}

marcgg 2010-01-06 14:55:11

Answer 4

+4 A:

Although detection of the sign of the product can be done, it's not what you are interested in. Especially if you're going to use it on large volumes of floats (eg to detect a zero crossing in a time stream).

The simplest way is to exactly express what you ask for: is the sign of a equal to the sign of b?

function samesign( a, b ) {
  var aPositive = a >= 0;
  var bPositive = b >= 0;
  return aPositive == bPositive;
}

Or shorter:

function samesign( a, b ) { return (a>=0) == (b>=0); }

xtofl 2010-01-06 15:21:27

+1 Better solution than a multiplication. As everyone notes in other answers, it is important to code clearly and without tricks. Unless you have other concers like optimization

Toad 2010-06-14 15:58:53

Answer 5

A:

Someone with almost the same question, but worried about +0 and -0

http://stackoverflow.com/questions/2922619/how-to-efficiently-compare-the-sign-of-two-floating-point-values-while-handling-n

Spudd86 2010-06-14 15:53:37

ansaurus

tags:

views:

answers:

comparing two floats to see if they're both negative, or both positive.

related questions