Quick question about the use of double

Question

0.00/5 (No votes)

See more:

Hi,

Im working on piece of code where the a calculation needs to be done, using the values entered by the user.

But strangely, the values are being approximated. For example if the user enters 28 and 32.8, the difference which should be 12.8 is calculated as 12.7777778.. But the kind of data I'm dealing with, needs precise values.

I have declared the variable as double.
Any suggestions????

Posted 28-Mar-12 5:01am

Sumal.V

Updated 28-Mar-12 21:51pm

v2

Add a Solution

Comments

Sergey Alexandrovich Kryukov 28-Mar-12 11:05am

Do you think that 100% accurate representation of real number is theoretically possible? It means that you don't know what is a real number...
--SA

Sumal.V 28-Mar-12 11:12am

Haha I do know that u cannot get 100% accuracy in calculations. But atleast the value entered by the user can match up! This will stop the accumulation of errors. And I work on latitudes and longitudes and slight change in seconds leads to many metres of difference!

Sergey Alexandrovich Kryukov 28-Mar-12 11:14am

Don't mix up real accuracy with string representation.
--SA

Richard MacCutchan 28-Mar-12 12:02pm

No it won't - except occasionally by accident. If you want precision use doubles; if you want accuracy use integers. Whether you like it or not, that is the only way in digital computers.

Stefan_Lang 28-Mar-12 11:08am

Hm, there used to be a series of Pentium CPUs with that kind of odd behaviour...
;-)

Ok, assuming it's not a CPU issue, it would be helpful to see both the code that reads the data and the code you use to display it. Without that it's really hard to tell.

Sumal.V 28-Mar-12 11:22am

Well its no complicated code: it looks like :

Class Box
{
double val;
}

class Book :: Set()
{
InsCtrl(80, EC_EDIT, b1.time1); // b1 is the object of Box.
}

class Scale
{
double delta = b1.time1 * 100 ; // the values here are approximated.

}

Stefan_Lang 28-Mar-12 11:28am

I don't see how that reads or displays a number.

Sumal.V 28-Mar-12 11:32am

Th set function in Book class, uses the Edit function, thats is used to provide an editable box and that value is used for calculation in Scale class in the delta variable .

Stefan_Lang 28-Mar-12 11:40am

Ok, so the user enteres the number via an edit box. He enters 28. Your Set function then continues to read that value, maybe convert to double, and store it in what in your code above would be b1.time1

That clarifies how you read the data. Now what did you do to find out that it's a different value, and what did your program do between the reading and the finding out?

Sumal.V 28-Mar-12 11:43am

Well I usually debug the code with the breakpoints set. That is how I know.

Sumal.V 28-Mar-12 11:49am

Oh sorry, the value is read the right way.
Actaully consider the double delta to be:
double delta = b1.time1 - b1.time2;

The user enters: time1= 1989, time2= 2011.8
but the answer for delta instead of 22.8 is 22.79999;

Stefan_Lang 28-Mar-12 11:58am

Ok, now that is quite different to just reading a whole number. The moment you got a fraction, you are no longer exact. That is exactly what SAKruykov told you.

A fraction of a number, even if easily represented as a decimal number, is not guranteed to be 100% correct when represented as a double value. That is what you're seeing. If the error you've noticed is bigger than 10^-12, then probably the problem is that your edit control converts the value to float, not double. If it's smaller, there's nothing you can do.

Sumal.V 28-Mar-12 12:04pm

Hmmm.. thanks anyways

Sergey Chepurin 29-Mar-12 5:44am

Read this first - "Floating point numbers": http://www.learncpp.com/cpp-tutorial/25-floating-point-numbers/
It will take 20-30 minutes, but can save you hours.

Sumal.V 29-Mar-12 7:49am

Thanks, that link is quite helpul

3 solutions

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Accepted Answer · 2012-03-28T05:13:00

Solution 1

Please see my comment. Did you ever heard that, for example, some infinite subset of the set of real numbers are transcendental numbers? The examples are π and e. Do you know that any representations of each of such numbers in the form of the string contains infinite number of digits. Suppose you have any arbitrary text of any finite length (say, Romeo and Juliette); code it in the form of, say, a string of decimal digits in any way. The fact is: in the decimal representation of, say π, one can find exact same sub-sequence. This is mathematically proven and well-know fact. Do you feel surprised?

This is a reasonable accuracy. Please see:
http://en.wikipedia.org/wiki/Double-precision_floating-point_format[^].

—SA

Posted 28-Mar-12 5:13am

Sergey Alexandrovich Kryukov

Comments

Sumal.V 28-Mar-12 12:04pm

thanks

Sergey Alexandrovich Kryukov 28-Mar-12 17:04pm

You are welcome.
If you agree that this makes sense, please consider accepting this answer formally (green button) -- thanks.
--SA

Espen Harlinn 28-Mar-12 15:58pm

5'ed!

Sergey Alexandrovich Kryukov 28-Mar-12 17:04pm

Thank you, Espen.
--SA

RaisKazi 28-Mar-12 20:45pm

My 5.

Sergey Alexandrovich Kryukov 28-Mar-12 21:16pm

Thank you, Rais.
--SA

ProEnggSoft 29-Mar-12 1:31am

+5

Sergey Alexandrovich Kryukov 29-Mar-12 10:52am

Thank you.
--SA

Espen Harlinn · Accepted Answer · 2012-03-28T05:37:00

Solution 2

Try using the GNU Multiple Precision Arithmetic Library [^].

The precision of GMP is only limited by available memory.

Best regards
Espen Harlinn

Posted 28-Mar-12 5:37am

Espen Harlinn

Comments

RaisKazi 28-Mar-12 20:46pm

My 5.

Espen Harlinn 5-Apr-12 5:10am

Thank you RaisKazi :-D

Abhinav S 5-Apr-12 4:56am

My 5.

Espen Harlinn 5-Apr-12 5:11am

Thank you Abhinav :-D

nv3 · Accepted Answer · 2012-03-28T10:00:00

I would not worry about precision too much for your application. If you are expressing latitude and longitude with C++ data type double on a normal Intel machine you will have at least 15 decimal digits of precision. An arc second represented as radian is 5 * 10**-6. So you have a precision of about 10**-9 arc seconds, each of which corresponds to 30m on the surface. So your results are precise to about 10**-7 meters, which is a tenth of a micrometer. That sounds like enough for most applications.

Another question is how to avoid all those 9s in results. 89.999999999 degrees latitude simply looks ugly. The secret to success here is rounding. When using printf, it will do the rounding for you if you specify a certain number of decimals, for example:

printf ("%.6f", 89.9999999999999);

will yield

90.000000

Now, if you convert your latitudes and longitudes to more human readable formats, like N53 34'47" then pay attention to implement the correct rounding mechanisms.

Quick question about the use of double

3 solutions

Solution 1

Solution 2

Solution 3

Add your solution here

Preview 0

Existing Members

...or Join us