[Cuis] Scoping rules for shared variables ?

[Peter van Rooijen]

Hi Juan, Hannes,

since you asked: the other Smaltalks I have looked at (Dolphin, VA, VW,
ST/X, VSE, IIRC) already do it as I want it to work
(Squeak does it "wrong" so perhaps it's not really a Smalltalk
implementation)

This is about the lookup logic when we see a name that is not a local name
(local to the block, method or object)

To resolve it (i.e., determine if it is a valid reference to a shared
variable), this is the order of searching for the name:

1 pool dictionaries declared in this class (from left to right I presume
but I don't care so I haven't paid attention)
2 classPool (i.e. the class variables defined in this class)
3 classPools in superclass chain
4 globals

The code in Squeak (and Cuis) does something recursive in the lookup logic
when it should not, which is probably based on an incorrect understanding
of the lookup rules. I have to admit that it is not easy to find how this
should be done, I have studied the Blue Book and ANSI-98 Smalltalk standard
(long long ago) and they were a bit fuzzy. But the logic all the other
dialects have also makes the most sense (to me) and works for some "clever"
stuff I do to make code 100% portable between dialects, so that's how I
like it to work.

Cheers, Peter

BTW Hannes, I'm not sure I would be able to change this in Dolphin, but in
the other dialects, probably. But like I said, there is no need, it is
Squeak (and Cuis) that's the odd one out.

On Tue, Jul 21, 2015 at 3:36 AM, Juan Vuletich <juan at jvuletich.org> wrote:

> On 7/19/2015 6:01 PM, H. Hirzel wrote:
>
>> On 7/19/15, Peter van Rooijen<peter at aba-instituut.nl>  wrote:
>> ....
>>   I
>>
>>> recently changed the scoping rules for shared variables (I think they're
>>> done wrong in the base image). I could change what I wanted easily. So I
>>> think I am not in some outsider author group. Instead I am happy that I
>>> could change what I wanted without needing anyone's approval or
>>> cooperation.
>>>
>>
> Great!
>
>  A note:
>> This applies to any version / distribution of Smalltalk. But it means
>> that you have your own version of the core system.
>>
>> Could you elaborate on the issue of the shared variables, maybe it is
>> something to include in the core Cuis image?
>>
>> --Hannes
>>
>
> Agree. It would be nice to know.
>
> Cheers,
> Juan Vuletich
>
> _______________________________________________
> Cuis mailing list
> Cuis at jvuletich.org
> http://jvuletich.org/mailman/listinfo/cuis_jvuletich.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://jvuletich.org/pipermail/cuis_jvuletich.org/attachments/20150721/62be7266/attachment-0004.html>