Representing an extremely small number in java?

[adam.balgach]

I have a requirement to represent a very very small number in a java
class of mine:


2.2204e-16


Does anyone have a straightforward solution to doing this? or am i
stuck with writing out something like

double _a = 0.0000000000000000....222204;

?

Thanks in advance,
A.

 

[mearvk]

Check out this API:
http://java.sun.com/j2se/1.5.0/docs/api/java/math/BigDecimal.html

I believe you can create arbitrarily small/precise decimals using an
implementation of this class.

mearvk

 

[Andreas Wollschlaeger]

I have a requirement to represent a very very small number in a java
class of mine:


2.2204e-16


Does anyone have a straightforward solution to doing this? or am i
stuck with writing out something like

double _a = 0.0000000000000000....222204;

?

Thanks in advance,
A.

How about

double a = 2.2204e-16d;

where the final 'd' makes sure it is a double literal; otherwise it
would be a 'real', implicitly casted to a 'double'!

Life can be that easy....

HTH
Andreas

 

[Mark Thornton]

How about

double a = 2.2204e-16d;

where the final 'd' makes sure it is a double literal; otherwise it
would be a 'real', implicitly casted to a 'double'!

You don't need the 'd' suffix --- floating point values are double by
default. So this suffices

double a = 2.2204e-16;

By contrast you do need the 'f' suffix here:

float a = 2.2204e-16f;


Mark Thornton

 

[EJP]

double a = 2.2204e-16d;

where the final 'd' makes sure it is a double literal; otherwise it
would be a 'real', implicitly casted to a 'double'!

What's a 'real'? The 'e' is enough to mark it as a double, you don't
need the 'd' as well.

 

[mearvk]

I should also add that since decimals are continuous (and the mapping
onto them is not), you only get a certain degree of accuracy using
Java's doubles. A double is 64 bits, which is quite alot. But remember,
there are an infinite number of decimals values between 0 and 1. So
when you do a comparison between two very small numbers which are
*similar*, the JVM may consider them to be equal, when in fact they are
not. This is where the BigDecimal implementation is a better solution.

May be germane, or not.

Mearvk