Hi guys, i got a simple problem of losing precision when trying to use a ratio against total number values.
first I'll get a ratio simply by dividing a number by a total value, then i will use that ratio to get the same percentage of another total.
So if I have 2 totals, say 100 and 200 and the number I want from my first total is 20. I will divide 20 by 100 to get my ratio, then times my second total by the ratio to get 40.
The problem lies in the method I got that uses this. all the numbers are double and I think the problem is that a double type isn't large enough to use for a ratio and loses precision.
For example if I use 465 and 930 as my totals and the number i want from my first total is 125. (if i do the same calculation) :
My ratio is 0.26881720430107525
and the amount I get from using this ratio against the second total is 249.99999999999997 when I should get 250.
I hope I have explained this well enough...
Is there a sure way to overcome this problem?
Another tip: if you are sure the solution is an integer number, then consider rewriting your code in order to do first the multiplication and dividing that by the second number:
First 125*930 = 116250 (use a int or long if nessecary)
116250 / 465 = 250
Each number system has exactly 10 different digits.
posted 10 years ago
thanks for the input guys.
the problem is it can be any number not just an integer.
I've had a look at the BigDecimal class but it doesn't seem to solve my problem but make it worse.
here is basically how I've done my code. It's not a practical example but shows my problem.
printout : ratio = 0.00215053763440860215
new total = 0.99999999999999999975
the new total should be the same as the original total. ie 1.
I can see where I am going wrong I just don't know how to fix this?
You are trying to represent a rational number as a digital number. Like you can never fully represent other rational numbers like 1/3 as a digital number (0.33333333333333333333...) you can never fully represent 1/465 as a digital number.
If you need to keep the full accuracy there is little else left but use a proper fractional class like Apache Commons Lang's Fraction. In short: