[r6rs-discuss] [Formal] Datum values, shared structure and eq? vs equal?

[Mikael Tillenius]

Michael Sperber wrote:
> Mikael Tillenius <mti at tillenius.com> writes:
>
>   
>> Description: The description of datum values in section 1.9 doesn't
>> forbidd shared structure. For example would the value of "(let ((x
>> (cons (cons 1 2) 3))) (set-cdr! x (car x)) x)" return a datum value? I
>> assume the answer i "no" but it would be nice to make it
>> explicit.
>>     
>
> This is the tutorial introduction, where `set-car!' isn't even
> mentioned.  
But we cannot ignore the fact that "set-car!", "vector-set!" and "eq?" 
are part of the languge.
> Moreover, this section is quite explicit (in the first two
> sentences) about the fact that no complete characterization is given here.
>   
I assume you mean the first two sentences of section 1. What confuses me 
is that I can find no definition of what a datum value is outside of 
subsection 1.9. I assume that the set of datum values is the set of 
values that can be represented by some syntactic datum. The value of "x" 
in my example above would in fact be a datum value, the same datum value 
that would be represented by "((1 . 2) . (1. 2))". Is this correct?

Ok, it all makes sense as far as datum values are concerned. I'm still 
not sure exactly when a "value" is something abstract and when it is 
something more concrete with an identity that can be tested with "eq?" I 
will reread the report with this in mind...

Btw, I really like the introduction of datum values. I think earlier 
revisions of the report was a bit unclear about if Scheme programs were 
sequences of characters (as implied by the grammar) or S-expressions (as 
implied by e.g. macros).

/Mikael