[plt-scheme] Please help test version 359.100

From: Jacob Matthews (jacobm at cs.uchicago.edu)
Date: Mon Nov 13 18:52:24 EST 2006

On Nov 13, 2006, at 5:31 PM, Robby Findler wrote:

> Is "some object" allowed to be multiple values or not? That text
> suggests not.

"Types are associated with values (also called objects) rather than  
with variables" (section 1.1) seems to be the only definition of what  
an 'object' is. I think it's clear that they are saying that values  
are also called objects, and that by the normal rules of English we  
can conclude that the singular 'object' means the same thing as the  
singular 'value'. So the question becomes, are multiple values the  
same thing as a single value?  The report seems to think not (I'm  
going here by the descriptions of the values and call-with-values  
functions in section 6.4), though it plays its usual trick of simply  
not defining what happens if a context expects a single value and it  
receives multiple ones.

Applying this back to the original question, I think we have to  
conclude that for-each is supposed to be able to be called in  
contexts that expect exactly one value. In Schemes like PLT Scheme  
where contexts that expect one value signal an error if they receive  
some other number, then it's a violation of R5RS for for-each to  
behave the way 359.100 did; but for Schemes like Bigloo that have  
coercion rules that take a multi-value return to a single-value  
context, it's not a violation.


(I actually wrote up a version of this before and then deleted it,  
thinking it was too technical for anybody to care about ... looks  
like I was wrong. :) )

Posted on the users mailing list.