"Rounding" bug incrementing a variable by 0.1 - KiXtart.org - official site

You are not logged in. [Log In] KiXtart.org website » Forums » KiXtart » Beta » "Rounding" bug incrementing a variable by 0.1

Page 1 of 1

1

Topic Options

#185117 - 2008-02-06 09:54 PM "Rounding" bug incrementing a variable by 0.1
Jochen Jochen KiX Supporter Registered: 2000-03-17 Posts: 6380 Loc: Stuttgart, Germany	Code: break on @kix ? ? $dbl = 0.0 while $dbl < 15 $dbl = 0.1 + $dbl $dbl ? loop get $ This one shows a whacky rounding behaviour both in 4.53 and 4.60 (both kix32 and wkix32) when it reaches 5.9 and adds 0.1 (no matter if $dbl is 0 or 0.0 or (not) implicitly/explicitly declared before the loop) Output on Windows XP Professional SP2 (German standard regional settings, therefore the comma in the output) : 4.60 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 1,1 1,2 ... normal output, until: 5,7 5,8 5,9 5,99999999999999 6,09999999999999 6,19999999999999 6,29999999999999 6,39999999999999 6,49999999999999 6,59999999999999 6,69999999999999 6,79999999999999 ... the bug persists until it suddenly goes back to normal at 10 9,79999999999998 9,89999999999998 9,99999999999998 10,1 10,2 10,3 10,4 10,5 _________________________
Top

#185126 - 2008-02-07 05:58 AM Re: "Rounding" bug incrementing a variable by 0.1 [Re: Jochen]
Sealeopard Sealeopard KiX Master Registered: 2001-04-25 Posts: 11165 Loc: Boston, MA, USA	Seen those things before, was typically due to an internal conversion of the number to floting point _________________________ There are two types of vessels, submarines and targets.
Top

#185128 - 2008-02-07 09:27 AM Re: "Rounding" bug incrementing a variable by 0.1 [Re: Sealeopard]
Richard H. Richard H. Administrator Registered: 2000-01-24 Posts: 4946 Loc: Leatherhead, Surrey, UK	Originally Posted By: Sealeopard Seen those things before, was typically due to an internal conversion of the number to floting point Agreed, it is not a bug. You will find this annoyance in common in all (most?) programming languages, and it is caused (as SL says) by the way that floating point numbers are stored, which is an approximation of the real value. There is a pretty good article here if you want more detail. I learned this the hard way back in the days i wrote 'C' code, but it applies equally to languages which are further away from the machine code. It's why when testing a floating point for zero you can never say Code: If $fpNumber = 0 Instead you have to use something like Code: If $fpNumber < 0.001 AND $fpNumber > -0.001 The values that you use for comparison will depend on the scale of your calculations.
Top

#185169 - 2008-02-08 03:05 PM Re: "Rounding" bug incrementing a variable by 0.1 [Re: Richard H.]
Jochen Jochen KiX Supporter Registered: 2000-03-17 Posts: 6380 Loc: Stuttgart, Germany	I would still call it a bug, not a KiXtart bug, but anyway a bug. Well, there might be a good explanation of things caused by approximation of things. I will read the article you linked Richard, and I will do this with great interest $\:\)$ But it still seems to be laziness to leave bugs, that may be unavoidable for complex math, as they are when it comes to simple additions. I mean, 5.9 + 0.1 is at any time 6.0 and not 5.99999999. Another strange fact is that there is no problem doing this increment with 0.2 or any higher number. Oh well, not quite true: As soon as the last digit of the floating point number is a 1 there will be sooner or later a rounding bug that "heals" itself shortly thereafter I see that there might be no easy fixing from Ruuds side (or any fixing at all), but still it smells fishy like the sloppy coding started in the sixties of last century that gave IT the great boost before the new millennium Edited by Jochen (2008-02-08 03:08 PM) _________________________
Top

Page 1 of 1

1

Previous Topic

Previous Topic View All Topics

View All Topics

Index Next Topic

Next Topic

Moderator: ShaneEP, Arend_, Jochen, Radimus, Glenn Barnas, Allen, Ruud van Velsen, Mart

Hop to:

Shout Box

Who's Online

0 registered and 1169 anonymous users online.

Newest Members

M_Moore, BeeEm, min_seow, Audio, Hoschi
17883 Registered Users

Generated in 0.059 seconds in which 0.03 seconds were spent on a total of 13 queries. Zlib compression enabled.