On 01/10/2008 01:15 PM, Martin Rubey wrote:
> Ralf Hemmecke <[EMAIL PROTECTED]> writes:
>
>> On 01/10/2008 09:42 AM, Martin Rubey wrote:
>>> Dear Gaby,
>>> I saw you implemented map!$Stack as follows:
>>> map!(f: S -> S, s: %) == -- from HOAGG
>>> map!(f, deref s)$List(S)
>>> s
>>> I have two questions: 1) why didn't you say
>>> map!(f: S -> S, s: %) == ref map!(f, deref s)$List(S)
>>> 2) isn't map!(f: S -> S, s: %) == -- from HOAGG
>>> map!(f, deref s)$List(S)
>>> s
>>> assuming that map! stores its result in s? I'm not sure whether this is
>>> ok:
>> Unfortunately 1) and 2) are not equivalent with the current implementation of
>> the Reference domain. In boolean.spad.pamphlet one finds
>>
>> Reference(S:Type): Type with
>> ref : S -> %
>> ++ ref(n) creates a pointer (reference) to the object n.
>> ...
>> == add
>> Rep := Record(value: S)
>> p = q == EQ(p, q)$Lisp
>> ref v == [v]
>> ...
>>
>> That looks to me like that for
>>
>> a: Reference Integer := ref 1;
>> b: Reference Integer := ref deref a;
>>
>> it does not hold
>>
>> a = b
>
> Quite possible. But I wouldn't need that in map! anyway, would I?
>
> the doc to map! only says
>
> map!(f,u) destructively replaces each element x of u by \axiom{f(x)}.
>
> and I think that should be interpreted as:
>
> the result of map!(f,u) is obtained by replacing each element x of u
> by
> f(x). In this process, u may be destroyed.
>
> I think that would allow the implementation
>
> map!(f: S -> S, s: %) == ref map!(f, deref s)$List(S)
>
> but *not*
>
> map!(f: S -> S, s: %) == -- from HOAGG
> map!(f, deref s)$List(S)
> s
>
>> That said ...
>>
>> *Don't use l after you have called sort!(l) unless you say l:=sort!(l).*
>
> That's exactly what I'm thinking. (for sort! it is obvious, but not that much
> for map!, that's why I asked)
Aha! Now I understand. And you are quite right.
map!(f: S -> S, s: %) == -- from HOAGG
map!(f, deref s)$List(S)
s
Is dangerous if it is not documented that this map!(f,u) leaves u in a
reasonable state. In fact, the documentation
if % has shallowlyMutable then
map_!: (S->S,%) -> %
++ map!(f,u) destructively replaces each element
++ x of u by \axiom{f(x)}.
in aggcat.spad seems to suggest a good u, but, in fact, I don't think
that it is the best API description.
So what about
map!(f: S -> S, s: %) == -- from HOAGG
setref(s, map!(f, deref s)$List(S))
s
?
With that definition we would have
s: Stack Integer := [1,2,3]
f: Integer -> Integer := ...
t := map!(f, s)
and s = t after the call, i.e. meaning that s is destructively modified,
but *usable* afterwards.
I now tend to be against 2) (the current implementation) since there
Stack silently relies on the *implementation* (not the interface) of
List. The interface of List *doesn't say explicitly* that the second
argument of map! is still usable after the call.
Ralf
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
open-axiom-devel mailing list
open-axiom-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/open-axiom-devel
