From: Jim Driscoll <j@...>
Date: 2004-02-08T19:46:11+09:00
Subject: Re: [NUBY]  Some simple (?) type questions

Hi,

On 8 Feb 2004, at 03:55, Silly wrote:

>
> First, I'm running into troubles with the Doubles. According to the 
> W3C, the "value space of double consists of the values m � 2^e, where 
> m is an integer whose absolute value is less than 2^53, and e is an 
> integer between -1075 and 970, inclusive." Now, I suspect that its 
> supposed to say "m � 10^e" but that isn't my problem. My problem is 
> that I simply can't seem to get the smallest values to come out as 
> anything other than 0. I'm sure its possible I'm just not sure where 
> to start.
>

Okay, apparently Ruby's floats are doubles, but that's probably not 
your problem. The formula is correct as-is, AFAIK, it's an expression 
of the number of bits available (in a 32-bit value) assigned to the 
(base-2) exponent. Anyway, I'd guess your only problem is:

irb(main):001:0> a=1/(2**970)
=> 0
irb(main):002:0> b=1.0/(2**970)
=> 1.00208418e-292

...how you're expressing creation of your very-small values. The 
numerator *must* be a float already, otherwise you're doing forced 
integer maths (Integer#/ not Float#/).

By the way, after poking around in IRB, it looks like ruby impliments 
64-bit floating point (if necessary), because I can make some extremely 
big numbers. Someone more knowledgeable can confirm if that's right.

> My second problem has to do with the way the W3 defines numbers in 
> general. In addition to numbers as we all understand, they've added 
> INF, -INF, NaN and -0. Each of these have very specific meanings, most 
> insteresting is that -0 < 0 and NaN > INF. Now I was thinking that I 
> would implement INFClass and NaNClass similar to TrueClass and 
> FalseClass. However, I have no idea how to handle -0.
>

All that stuff is built right into the design of the floating point 
standard (It's an IEEE standard. Google for it), which ruby impliments:

irb(main):001:0> a=1.0
=> 1.0
irb(main):002:0> a/=0
=> Infinity
irb(main):003:0> b=0.0
=> 0.0
irb(main):004:0> b/=0
=> NaN
irb(main):005:0> c=-1.0
=> -1.0
irb(main):006:0> c/=0
=> -Infinity
irb(main):007:0> d=1
=> 1
irb(main):008:0> d/=c
=> -0.0
irb(main):009:0>

So ruby handles it all for you (even if it doesn't seem to have 
constants for the more 'interesting' values). See Float#infinity? and 
Float#nan? too.


Jim