Re: [Jscheme-user] Question about exact numbers

[Timothy H. <tim...@ma...>]

Hi Nate,

I would agree with Ken's assessment here.

If you want to provide your users with a full R5RS implementation as an 
applet
and if you can assume they have Java 1.2 enabled browsers, then you can 
use
the recently released SISC applet

      http://sisc.sourceforge.net/sisc-online.html

The advantage of Jscheme over SISC is the tighter integration with Java 
and hence
the ability to easily write Scheme libraries that provide access to the 
corresponding
Java libraries. This can be done is SISC, but the interface is less 
natural as you need
to manually convert between Scheme types and Java types at every Java 
call.

If you have a new student working on the project, you might be able to 
have them
connect Darrel's editor with the SISC applet, getting the best of both 
worlds....

Nate, you might also be interested in some of the open source
groupware tools we are developing for
use in teaching intro Scheme and Java classes. An early prototype is 
available under
the TATool link from my page (http://www.cs.brandeis.edu/~tim) or you 
can get it
directly at
    http://frege.cs-i.brandeis.edu:8090/user/tatool/tatoolv1a.sxml
but this assumes you have java web start installed
   http://java.sun.com/products/javawebstart


Hmmmm.

<snip comment="cut here to remove developer musings.....">

Ken, We ought to be able to write a Jscheme interface to the SISC 
interpreter!
The Jscheme applet would provide the GUI, paren matching, window 
management,
network interface, groupware access, etc. and
it could call the SISC interpreter to evaluate the R5RS Scheme 
expressions....

Perhaps we can provide a simple interface between Jscheme and SISC.....
to make this painless....
</snip>

Anyway, if there is anything I can do to help, don't hesitate to ask.
---Tim---


On Thursday, May 30, 2002, at 08:57 PM, Ken Anderson wrote:

> When we started in 1998, our implementation of scheme in Java was 
> pretty close to Scheme.  We steadily moved closer to Java.  This has 
> given us easy access to Java API's, to the extent that Jscheme is 
> really a Scheme alternative syntax to Java.  We get deep access to Java 
> with a relatively small amount of code.  For example, (define now 
> (Date.))
>
> The disadvantage is, as you pointed out, that if you want a full bignum 
> number tower, to verify some serious computation, for example, you have 
> to look elsewhere.  Jscheme, don't provide anything more than Java 
> does, so (/ 1 2) is 0.  More generally, Scheme people, who might be 
> attracted to us at first, might feel discouraged by the Java focus.  
> This is certainly an issue we're concerned about and would like to hear 
> people's views.
>
>
> In PLT Scheme, (/ 1 2) is 0.5, as you expect.
>
> Java does have bignum packages, and we've talked them on this list.  An 
> extension to Jscheme that provides such support should not be hard to 
> write.
> If you're interested, i'd be willing to help.
>
> The http://sisc.sourceforge.net scheme is a full rsr5 scheme 
> implementation in java.  At this point, it is harder to interface to 
> Java than Jscheme is.  It is also faster on some benchmarks, though 
> Jscheme does suprisingly well.
>
> k
>
> At 07:42 PM 5/30/2002, Ken Anderson wrote:
>> At 06:40 PM 5/30/2002, Nathaniel Titterton wrote:
>>
>>> (I've posted this on the open discussion on the sourceforge site, but 
>>> read
>>> that I should try here as well).
>>>
>>> I was playing with a random number generator, and was surpised to find
>>> that in JScheme
>>>
>>> (/ 1 2)
>>>
>>> was returning 0. If jscheme doesn't have exact rational numbers (I'm
>>> assuming it doesn't), wouldn't it make more sense to return 0.5 
>>> instead?
>>> An implemenation is free to silently do this, it says in R4RS.
>>
>> We did this in an earlier version of Jscheme, then called Silk.
>> We now adapt Java's arithmetic model so (/ 1 2) is 0
>> and (/ 1 2.0) is 0.5.
>>
>> Section 6.5.3 of R4RS gives some leeway in how numbers are implemented.
>>
>>> -nate
>>>
>>>   
>>> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>>>   Nathaniel Titterton                        
>>> na...@so...
>>>   CITRIS PostDoc       341 Soda Hall, U.C. Berkeley, Berkeley, CA 
>>> 94720
>>>
>>>
>>> _______________________________________________________________
>>>
>>> Don't miss the 2002 Sprint PCS Application Developer's Conference
>>> August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
>>>
>>> _______________________________________________
>>> Jscheme-user mailing list
>>> Jsc...@li...
>>> https://lists.sourceforge.net/lists/listinfo/jscheme-user
>>
>>
>> _______________________________________________________________
>>
>> Don't miss the 2002 Sprint PCS Application Developer's Conference
>> August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
>>
>> _______________________________________________
>> Jscheme-user mailing list
>> Jsc...@li...
>> https://lists.sourceforge.net/lists/listinfo/jscheme-user
>>
>
>
> _______________________________________________________________
>
> Don't miss the 2002 Sprint PCS Application Developer's Conference
> August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
>
> _______________________________________________
> Jscheme-user mailing list
> Jsc...@li...
> https://lists.sourceforge.net/lists/listinfo/jscheme-user
>