Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.prolog > #14816

Electing "conductor" Prolog variables (Was: Do Prologs Dream of Canonical Sheep?)

Show all headers | View raw

Hi,

This is applicable to certain Prolog systems since
in Prolog without change_arg/3 loops can ideeded only
go through Prolog variables, making the Prolog

variables the conductor elements. It won’t work so
easily when a “mutator” like change_arg/3 is available,
that bypasses Prolog variables. And also if we don’t

have change_arg/3 the hash consing might be less
effective if we consider all Prolog variables. How
to make a smart and efficient selection of the

right “conductor” elements?

Bye

BTW: There is also some overlap to WebPL and their
recent Heap/Stack architecture and some insuination
about variable shunting. WebPL seems to be still active,

currently trying Scryer chars.

Mild Shock schrieb:
> Hi,
> 
> @jp-diegidio repeatedly put forward a Prolog
> system vision, that does Prolog term minimization
> “intrinsically”. Such novel design ideas, just
> 
> like E-graphs by Philip Zucker, this CanProlog
> by Julio Di Egidio, is probably best discussed in
> a new thread. And here you have it, a new thread!
> 
> BTW, I think there is some overlap of E-graphs and
> CanProlog. For example was researching whether
> Hash Consing is possible for cyclic terms. There
> 
> is indeed an old proposal by Andre W. Appel from
> Princton University, burried in this paper:
> 
> Hash-consing Garbage Collection
> Appel, A.W. - 1993
> https://www.cs.princeton.edu/~appel/papers/hashgc.pdf
> 
> The idea is simple just have some “conductor”
> elements, where hash consing stops. And loops can go
> through these “conductor” elements.
> 
> Bye
> 
> Mild Shock schrieb:
>> Hi,
>>
>> Lee Naish’s idea for an infinite tree syntax:
>>
>> /* Lee Naish’s idea */
>> naish(X, Y) :-
>>     naish([], X, Y).
>>
>> naish(S, X, Y) :- compound(X),
>>     member(Z-Y, S), X == Z, !.
>> naish(S, X, Y) :- compound(X), !,
>>     length(S, N),
>>     X =.. [F|L],
>>     maplist(naish([X-'$IDX'(N)|S]), L, R),
>>     Y =.. [F|R].
>> naish(_, X, X).
>>
>> Now define a compare:
>>
>> naish_compare(C, X, Y) :-
>>     naish(X, A),
>>     naish(Y, B),
>>     compare(C, A, B).
>>
>> Works fine on this example:
>>
>> ?- X = s(s(1,Y),0), Y = s(s(1,X),1),
>>     naish_compare(C, X, Y).
>> C = (>).
>>
>> The residual problem is, that it is extremly slow. Much slower
>> than the native compare/3 which doesn’t use a stack, but a Union
>> Find structure. So conceptually total order and natural order
>>
>> compare/3 are easy, the hard part is making them ultra fast!
>>
>> Bye
> 

Back to comp.lang.prolog | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

The Million Dollar question of Prolog Mild Shock <janburse@fastmail.fm> - 2025-08-23 14:54 +0200
  AGI will relieve the community (Was: The Million Dollar question of Prolog) Mild Shock <janburse@fastmail.fm> - 2025-08-25 15:11 +0200
    Reflect on Sience not Frog Legs (Was: AGI will relieve the community) Mild Shock <janburse@fastmail.fm> - 2025-08-25 22:55 +0200
      Switching Gears: From Lee Naish to Peter Deutsch (Was: Reflect on Sience not Frog Legs) Mild Shock <janburse@fastmail.fm> - 2025-08-26 11:24 +0200
        Not for pratical use, only for educational use (Re: Switching Gears: From Lee Naish to Peter Deutsch) Mild Shock <janburse@fastmail.fm> - 2025-08-26 11:32 +0200
  Do Prologs Dream of Canonical Sheep? (Was: The Million Dollar question of Prolog) Mild Shock <janburse@fastmail.fm> - 2025-09-05 12:10 +0100
    Electing "conductor" Prolog variables (Was: Do Prologs Dream of Canonical Sheep?) Mild Shock <janburse@fastmail.fm> - 2025-09-05 12:13 +0100

csiph-web