In C#: double d = 0.4 - 1.2; What is d?

[Ivor S Sargoytchev]

Chris, With respect, some constructive criticism please. Ivor S. Sargoytchev Dundas Software

[peterchen]

ok, you asked for it... It is *well known* (;P) that floating point numbers have a limited resolution (around 7 digits for single, 15 for double*), and comparing the result of an arithmetic operation using "==" is prone to err. if you know the order of magnitude of one comparand, or at least you know that it's "far enough" from zero, use the abs(a-b) < 1e-6*(abs of known comparand) comparison. Otherwise it gets a little bit more tricky. *) don't quote me on the numbers, look up the IEEE definitions for the format

---

"Der Geist des Kriegers ist erwacht / Ich hab die Macht" StS
sighist | Agile Programming | doxygen

[Chris Maunder]

The value represented in a floating point variable is only an approximation to a value, not an exact representation. Because of this you cannot assume that something like 2.0 == 4.0 * 0.5 will return true. When comparing floating point values you need to estimate how close is close enough then do a fuzzy comparision (something like ABS(A-B) < Epsilon) cheers, Chris Maunder

[Lost User]

Chris Maunder wrote: The value represented in a floating point variable is only an approximation to a value, not an exact representation. Because of this you cannot assume that something like 2.0 == 4.0 * 0.5 will return true. So this is how you count your beer consumption isn't? Especially when Kiwi and Aussie mates are involved. ;P Michael Martin Australia "I personally love it because I can get as down and dirty as I want on the backend, while also being able to dabble with fun scripting and presentation games on the front end." - Chris Maunder 15/07/2002

[Brit]

If it's anything like C++, then d = -0.79999999999999993 ------------------------------------------ The ousted but stubbornly non-dead leader reportedly released an audiotape this weekend, ending by calling on Iraqis to, quote, "resist the occupation in any way you can, from writing on walls, to boycotting, to demonstrating and taking up arms." adding, "you know, pretty much anything I used to kill you for."

[Michael Dunn]

Chris Maunder wrote: 2.0 == 4.0 * 0.5 Won't that actually be true because all those numbers are powers of 2? :cool: --Mike-- Ericahist | Homepage | RightClick-Encrypt | 1ClickPicGrabber Latest art~!@#2rDFA#@(#*%$Rfa39f3fqwf--= NO CARRIER

[Michael Dunn]

That reminds me of the old C= BASIC, there was some number (I'll just use 42 as an example) where you could type:

PRINT 42

and the response would be:

42.0001

:wtf: --Mike-- Ericahist | Homepage | RightClick-Encrypt | 1ClickPicGrabber Latest art~!@#2rDFA#@(#*%$Rfa39f3fqwf--= NO CARRIER

[John M Drescher]

Yes it should... John

[Navin]

Ivor S. Sargoytchev wrote: Dundas Software THe only way you're going to get an answer in this crowd, when posting in the Lounge, is if you offer to hook us up with those women from the Summer Sale ads. :-D "When a man sits with a pretty girl for an hour, it seems like a minute. But let him sit on a hot stove for a minute and it's longer than any hour. That's relativity." - Albert Einstein

[Rodolfo Lima]

I think that what he meant was something like: if(abs(0.4-1.2)-0.8) < Double.Epsilon) which is the proper way to compare floating point numbers, i guess.

[John M Drescher]

I agree.. :):laugh: John

[Ivor S Sargoytchev]

This cannot be true since the absolute value of ((0.4 - 1.2) + 0.8) is 0.00000000000000011102230246251565. The value of Double.Epsilon is 4.94065645841247E-324, which is much much smaller.

[Ryan Binns]

Michael Dunn wrote: Won't that actually be true because all those numbers are powers of 2? Yeah it will. Not a good choice of numbers huh? :)

Ryan

"Punctuality is only a virtue for those who aren't smart enough to think of good excuses for being late" John Nichol "Point Of Impact"

[Ivor S Sargoytchev]

So Epsilon should be some value that you pick up, not Double.Epsilon as some people suggested, right? P.S. Chris, I know you are not supposed to compare 2 doubles with equality test. I was making a point that the result was not -0.8 as a response to leppie who wrote: "Thru a microscope, everything looks big. Believe me that is -0.8. Try Console.Write()." So, I accept your apology.:) Ivor S. Sargoytchev Dundas Software

[igor1960]

I understand that this thread is not serious about anything, but very valid point has been raised. I would give you an answer: nobody knows... until native code is created. During days when C ruled you could heave an option of switching on/off coprocessor, floating point emulator library and etc. So, you can adjust precissions and etc... Now, looks like JIT should be creating best possible code depending on your computer configuration: what exactly does that mean for floating point case -- I didn't figure out yet. But yes, they will be alot of knowledgeable people on this thread willing to answer -- no doubt... "...Ability to type is not enough to become a Programmer. Unless you type in VB. But then again you have to type really fast..." Me

[Maxwell Chen]

In Visual C++ 6.0, d = -0.80000004172325 ;P BuggyMax

[Nick Parker]

Ivor S. Sargoytchev wrote: So Epsilon should be some value that you pick up, not Double.Epsilon as some people suggested, right? Nope, check your local MSDN documentation on the Double.Epsilon Field. .NET Documentation wrote: For example, the C# expression, (double)1/3 == (double)0.33333, does not compare equal because the division operation on the left-hand side has maximum precision while the constant on the right-hand side is only precise to the visible digits. Instead, determine if the two sides of a comparison are close enough to equal for your purposes by comparing whether the absolute value of the difference between the left and right-hand sides is less than Epsilon. -Nick Parker

[Stephane Rodriguez]

Ok, so MS has still problem defining a simple and safe addition/substract operation on floating points. Nice. Now that's high-tech. Where is the Accessories / calculator SDK when I need it ? :laugh: Taking advantage of InternetExplorer to steal user's name and password. Taking advantage of InternetExplorer to steal user's clipboard.

[Olli]

Nemanja Trifunovic wrote: Hehehe, I ran into this problem 15 years ago while writing a Fortran program. AAAAAAAAAAAAAAAAARRRRRRRRRRRRRRRGHHHHHHHHH You made my old trauma come back.... Fortrandoubleintc++verystrangehelpfirstclass :wtf::wtf::wtf::wtf::omg::omg:

Olli "Ooooooh, they have the internet on computers now!"
Homer Simpson
:beer: + :java: = NULL :=> X|

[Brit]

BuggyMax wrote: In Visual C++ 6.0, d = -0.80000004172325 How'd you come up with that answer? Because I came up with a different number with Visual Studio 6.0: -0.79999999999999993 BTW, you'd better not say that you did "d += 0.8" and then looked at the value in the debugger. ------------------------------------------ The ousted but stubbornly non-dead leader reportedly released an audiotape this weekend, ending by calling on Iraqis to, quote, "resist the occupation in any way you can, from writing on walls, to boycotting, to demonstrating and taking up arms." adding, "you know, pretty much anything I used to kill you for."