Re: [Chicken-users] update: complex egg, problems with SRFI-10 implementation

[Thomas Chust]

Felix Winkelmann wrote:
> TiM wrote:
>
> > greetings programs.
> >
> > i have a few questions about the complex egg.
> >
> > first, wouldn't it be slightly more efficient to represent a complex
> > number as a length 2 vector?
> >
> > secondly, don't we need some sort of type identifier to distinguish a
> > complex number from a pair, or a length 2 vector?
> >
> > i'm thinking of something like ('complex . (a . b)) or ('complex . #(a 
> > b)).
>
> Actually a seperate data type would be best (could just be a custom
> record type with an accompanying printer).
> [...]

Hello,

If you think that would be best, why shouldn't I do it this way ;) A new 
version using a SRFI-9 record type instead of conses is available from 
the usual location.

I also defined a SRFI-10 reader constructor to go along with the record 
printer and ran into the following problem:

Suppose you do this in csi:

#;4> (define-record blubb a b)
#;5> (define-reader-ctor 'blubb make-blubb)

Then

#;6> #,(blubb 1 2)
Error: illegal non-atomic object: #<blubb>

yields an error -- but why?

This works, of course:

#;6> (make-blubb 1 2)
#<blubb>

Any ideas where this strange behaviour comes from?

cu,
Thomas