A problem with decimals...

[Cristian Martinello]

Try this code, and tell me how can I fix it, please.

var num1=new Number(parseFlo at("118.18"))
var num2=new Number(parseFlo at("50"))
var num3=new Number(parseFlo at("50"))
alert(num1-num2-num3)

This is very strange...


--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG

[Jerry Park]

Re: A problem with decimals...

Cristian Martinello wrote:[color=blue]
> Try this code, and tell me how can I fix it, please.
>
> var num1=new Number(parseFlo at("118.18"))
> var num2=new Number(parseFlo at("50"))
> var num3=new Number(parseFlo at("50"))
> alert(num1-num2-num3)
>
> This is very strange...
>
>[/color]
No its not. Javascript, like most programming languages, use floating
point arithmetic. Fractional numbers which are finite in one base may be
'repeating decimals' in another base. Floating point converts the
numbers to base 2, does the arithmetic, then converts the answer to base 10.

Rounding the answer to the precision of the operands is usually
sufficient. If not, convert the numbers to integers, perform the math,
then convert back.

[Lasse Reichstein Nielsen]

Re: A problem with decimals...

"Cristian Martinello" <cattivofics@ti scali.it> writes:
[color=blue]
> Try this code, and tell me how can I fix it, please.
>
> var num1=new Number(parseFlo at("118.18"))
> var num2=new Number(parseFlo at("50"))
> var num3=new Number(parseFlo at("50"))
> alert(num1-num2-num3)[/color]

Just
alert(118.18-100)
will do the trick.
[color=blue]
> This is very strange...[/color]

This is completely normal when calculating with finite precission
binary-based floating point numbers.

The computer cannot represent 118.18 exactly. It finds a
representation that is close enough that when output, it is rounded to
118.18. Now you subtract 100 from it. That means that the integer part
of the number gets smaller, so there are more bits available for
the fractional part.

Imagine the number is stored as 15 decimal digits, but not exactly:
118.18000000000 1
When you output it, the least significant digit is rounded away.
Now subtract 100 and still use 15 decimal digits:
18.180000000001 0
Suddently the error on the least significant bit is amplified, so
it won't go away by rounding.
The same problem happens with binary representation, typically when
making the number smaller.

<URL:http://jibbering.com/faq/#FAQ4_7>

/L
--
Lasse Reichstein Nielsen - lrn@hotpop.com
Art D'HTML: <URL:http://www.infimum.dk/HTML/randomArtSplit. html>
'Faith without judgement merely degrades the spirit divine.'

[Richard Cornford]

Re: A problem with decimals...

"Cristian Martinello" <cattivofics@ti scali.it> wrote in message
news:7d955b5e65 d9821c97998b092 db6736c.119552@ mygate.mailgate .org...[color=blue]
> Try this code, and tell me how can I fix it, please.
>
> var num1=new Number(parseFlo at("118.18"))
> var num2=new Number(parseFlo at("50"))
> var num3=new Number(parseFlo at("50"))
> alert(num1-num2-num3)
>
> This is very strange...[/color]

It is not strange, it is normal. It is a manifestation of the inability
of an IEEE double precision floating point number (as used by JavaScript
for its one number type) to accurately represent all fractions. What you
are seeing is the nearest approximation of 18.18 that the number format
can accommodate. It is not a problem unique to computer floating point
number formats, how would you write 1/3 as a decimal fraction?
(precisely?)

<URL: http://www.jibbering.com/faq/#FAQ4_7 >

What is strange is wrapping a - new Number - call around a -
parseFloat - call. You would get the same results from:-

var num1="118.18"
var num2="50"
var num3="50"
alert(num1-num2-num3)

- as the subtraction operator will force the required string to number
type conversion.

Richard.

[TrueBlue]

Re: A problem with decimals...

A typical trick to solve that is the following:

var result=Math.rou nd((num1-num2-num3)*10000)/10000;
alert(result)

that is, you round something that you also m,ultiplicate by some power of
ten 8depending on how many floats you want to keep) and then you divide
exactly by that same amount of power of ten out of the rounding process -
that fixes it in a way that many use.

http://www.unitedscripters.com/

"Cristian Martinello" <cattivofics@ti scali.it> ha scritto nel messaggio
news:7d955b5e65 d9821c97998b092 db6736c.119552@ mygate.mailgate .org...[color=blue]
> Try this code, and tell me how can I fix it, please.
>
> var num1=new Number(parseFlo at("118.18"))
> var num2=new Number(parseFlo at("50"))
> var num3=new Number(parseFlo at("50"))
> alert(num1-num2-num3)
>
> This is very strange...
>
>
> --
> Posted via Mailgate.ORG Server - http://www.Mailgate.ORG[/color]

[Thomas 'PointedEars' Lahn]

Re: A problem with decimals...

TrueBlue wrote:[color=blue]
> A typical trick to solve that is the following:
>
> var result=Math.rou nd((num1-num2-num3)*10000)/10000;
> alert(result)
>
> that is, you round something that you also m,ultiplicate by some power of
> ten 8depending on how many floats you want to keep) and then you divide
> exactly by that same amount of power of ten out of the rounding process -
> that fixes it in a way that many use.[/color]

Where's the point? Math.round(num1 - num2 - num3) will also do the trick.
[color=blue]
> [Fullquote below the text][/color]

Please don't do this, bandwidth is precious.


PointedEars