[racket] low level question: why begin0?

I think the point of 'begin0' is that it handles multiple values 
efficiently (ie, more so than a heap-allocated list). The expansion of 
'begin0' would have to be something like

(call-with-values (lambda () e1)
   (lambda result (begin e2 (apply values result))))

since in general the number of values produced by e1 is not known at 
expansion time.

Ryan


On 04/29/2012 03:54 PM, Sam Tobin-Hochstadt wrote:
> Presumably we could remove it from fully-expanded code and then the
> compiler could re-transform this:
>
> (let ([x e]) e0 ... x)
>
> into some bytecode that uses begin0, which would simplify the job of
> tools that process expanded code without changing performance.
>
> On Sun, Apr 29, 2012 at 5:07 PM, Robby Findler
> <robby at eecs.northwestern.edu>  wrote:
>> We debated this long ago and I think the conclusion was that we could
>> get a little more performance by including it at the bytecode level
>> that seemed worth having (the precise details (like if we had a
>> program where that mattered) escape me).
>>
>> Robby
>>
>> On Sun, Apr 29, 2012 at 1:03 PM, Danny Yoo<dyoo at cs.wpi.edu>  wrote:
>>> One question that I had stowed away a long time ago: why is begin0
>>> part of the language?  It seems redundant in the face of having 'let'
>>> to capture a value that we want to return at the end of some sequence.
>>> ____________________
>>>   Racket Users list:
>>>   http://lists.racket-lang.org/users
>>
>> ____________________
>>   Racket Users list:
>>   http://lists.racket-lang.org/users
>
>
>