Re: (real? (* +i +i)) -> #f

[Mark H Weaver <mhw@netris.org>]

On 08/03/2012 06:02 AM, JihemD wrote:
> I am playing around with Guile 2.05 on Kubuntu 12.04,
> why :
> scheme@(guile-user)>  (real? (* +i +i))
> $13 = #f
> but :
> scheme@(guile-user)>  (zero? (imag-part (* +i +i)))
> $14 = #t

In recent Scheme standards (both R6RS and draft R7RS), a number is 
considered real if and only if the imaginary part is an _exact_ zero. 
See the definitions of 'real?' in R6RS section 11.7.4 and R7RS section 
6.2.6.  The R6RS Rationale document provides some discussion in the 
"Flow analysis" portion of section 11.6.6.

The R6RS provides a 'real-valued?' predicate that might be closer to 
what you expect.  You can get it in Guile by importing (rnrs base).

As Ian noted, Guile does not currently support exact non-real complex 
numbers, so +i cannot be represented exactly, and thus the imaginary 
part of +i^2 is not known to be exactly zero.  If we add support for 
exact non-real complex numbers some day, then (real? (* +i +i)) will 
become #t.

> Thx Ian, the concept of exactness is new for me

The concept of exact numbers is described in R6RS section 3.2.

      Mark