Why 'inexact' and 'exact' doesn't check 'number?' first?

Why 'inexact' and 'exact' doesn't check 'number?' first?

[Nala Ginrut]

It's weird to see that:
(exact? 'a) 
================err msg===============
ERROR: In procedure exact?:
ERROR: In procedure exact?: Wrong type argument in position 1: a
==================end=================

And I have to do this:
(define (fraction? obj)
  (and (number? obj) (inexact? obj)))


Why not 'exact?' and 'inexact?' doesn't check 'number?' first? 

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Daniel Hartwig]

On 12 December 2012 11:21, Nala Ginrut <nalaginrut@gmail.com> wrote:
> It's weird to see that:
> (exact? 'a)
> ================err msg===============
> ERROR: In procedure exact?:
> ERROR: In procedure exact?: Wrong type argument in position 1: a
> ==================end=================
>
> And I have to do this:
> (define (fraction? obj)
>   (and (number? obj) (inexact? obj)))

“Fraction” is not a very precise term or relevent to the type
hierarchy in Scheme, please do not use that.  Disregard that GOOPS
define such a class, the more precise term is “rational” for which
there is already a predicate.

Your fraction? predicate is not equivalent to the GOOPS <fraction>
class, which are exact? and (in all cases I believe) also rational?.

>
>
> Why not 'exact?' and 'inexact?' doesn't check 'number?' first?

Performance. They are intended to be used in numerical paths and
requiring them to check first number? every time is very inefficient.
You will note that they are documented as “exact? Z“ rather than
“exact? obj”.

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Nala Ginrut]

On Wed, 2012-12-12 at 12:39 +0800, Daniel Hartwig wrote:
> On 12 December 2012 11:21, Nala Ginrut <nalaginrut@gmail.com> wrote:
> > It's weird to see that:
> > (exact? 'a)
> > ================err msg===============
> > ERROR: In procedure exact?:
> > ERROR: In procedure exact?: Wrong type argument in position 1: a
> > ==================end=================
> >
> > And I have to do this:
> > (define (fraction? obj)
> >   (and (number? obj) (inexact? obj)))
> 
> “Fraction” is not a very precise term or relevent to the type
> hierarchy in Scheme, please do not use that.  Disregard that GOOPS
> define such a class, the more precise term is “rational” for which
> there is already a predicate.

Many thanks!
But rational? returns TRUE both 1.5 & 1/2, I need a predicate to check
between 'float' and 'fraction' distinctly.

Are you suggesting I use (is-a? obj <fraction>) for 'fraction?' ?



> 
> Your fraction? predicate is not equivalent to the GOOPS <fraction>
> class, which are exact? and (in all cases I believe) also rational?.
> 
> >
> >
> > Why not 'exact?' and 'inexact?' doesn't check 'number?' first?
> 
> Performance. They are intended to be used in numerical paths and
> requiring them to check first number? every time is very inefficient.
> You will note that they are documented as “exact? Z“ rather than
> “exact? obj”.

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Daniel Hartwig]

On 12 December 2012 13:55, Nala Ginrut <nalaginrut@gmail.com> wrote:
> Are you suggesting I use (is-a? obj <fraction>) for 'fraction?' ?

Absolutely not.  Use inexact? if you wish to determine that the
*storage* of a value is using floating point format.

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Mark H Weaver]

Daniel Hartwig <mandyke@gmail.com> writes:
> On 12 December 2012 13:55, Nala Ginrut <nalaginrut@gmail.com> wrote:
>> Are you suggesting I use (is-a? obj <fraction>) for 'fraction?' ?
>
> Absolutely not.  Use inexact? if you wish to determine that the
> *storage* of a value is using floating point format.

Apologies in advance for being pedantic, but there is no guarantee that
inexact numbers are represented in floating point format.  Having said
that, I'm not aware of any current Scheme implementation that uses a
different representation for inexacts.

Nala Ginrut <nalaginrut@gmail.com> writes:
> (define (fraction? obj)
>   (and (number? obj) (inexact? obj)))

This definition implies that 3.1415927 is a fraction, and 1/2 is not.
I suspect you want something closer to this:

  (define (fraction? obj)
    (and (rational? obj)
         (exact? obj)
         (not (integer? obj))))

     Regards,
       Mark

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Daniel Hartwig]

On 12 December 2012 14:32, Mark H Weaver <mhw@netris.org> wrote:
> Daniel Hartwig <mandyke@gmail.com> writes:
>> On 12 December 2012 13:55, Nala Ginrut <nalaginrut@gmail.com> wrote:
>>> Are you suggesting I use (is-a? obj <fraction>) for 'fraction?' ?
>>
>> Absolutely not.  Use inexact? if you wish to determine that the
>> *storage* of a value is using floating point format.
>
> Apologies in advance for being pedantic, but there is no guarantee that
> inexact numbers are represented in floating point format.  Having said
> that, I'm not aware of any current Scheme implementation that uses a
> different representation for inexacts.

Quite right.

Pedantism is appropriate given the context of writing a Scheme
expression printer.

Re: Why 'inexact' and 'exact' doesn't check 'number?' first?

[Andy Wingo]

On Wed 12 Dec 2012 04:21, Nala Ginrut <nalaginrut@gmail.com> writes:

> It's weird to see that:
> (exact? 'a) 
> ================err msg===============
> ERROR: In procedure exact?:
> ERROR: In procedure exact?: Wrong type argument in position 1: a
> ==================end=================

scheme@(guile-user)> (positive? "foo")
ERROR: In procedure positive?:
ERROR: In procedure positive?: Wrong type argument in position 1: "foo"

Same reason; exact? operates on numbers.

Cheers,

Andy
-- 
http://wingolog.org/