Path: ...!weretis.net!feeder8.news.weretis.net!reader5.news.weretis.net!news.solani.org!.POSTED!not-for-mail
From: Mild Shock <janburse@fastmail.fm>
Newsgroups: comp.lang.prolog
Subject: Side Note, Why not DSLs? (Re: Chicken and egg, with curry?)
Date: Fri, 3 Jan 2025 22:09:47 +0100
Message-ID: <vl9jmr$25sjr$2@solani.org>
References: <vl9fs3$3ui38$1@dont-email.me> <vl9hq8$3ui38$2@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Fri, 3 Jan 2025 21:09:47 -0000 (UTC)
Injection-Info: solani.org;
	logging-data="2290299"; mail-complaints-to="abuse@news.solani.org"
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Firefox/91.0 SeaMonkey/2.53.19
Cancel-Lock: sha1:5nSGiBGk5Ren0aJG5eY1AhV/+fw=
X-User-ID: eJwNysERACEIA8CWREg4y1EJ/Zdwzn4XTuPNIBjoZ2VLww+4dSgFJPdZwJ2tOG67a3298X5Vpg1SnAKnlf90GhYA
In-Reply-To: <vl9hq8$3ui38$2@dont-email.me>
Bytes: 5484
Lines: 147

Hi,

Just a side note, don't let you get distracted.
A side node about the chicken/egg problem, i.e.
without the curry:

- a) Eggs came first, for example Turtles had eggs
      Turtles are part of an ancient reptilian
      lineage ca 300 million years ago

- b) Chickens came after Turtles
      Chickens, on the other hand, are much younger
      in comparison, evolved from theropod dinosaurs
      around 150 million years ago

Not sure whether this helps. But I think it could help
nevertheless:

- i) Logic Programming is the Egg

- ii) From the Egg Turtles or Chickens can hatch,
     its very easy to program functionally or
     procdurally in Prolog. Just add small DSLs:

https://en.wikipedia.org/wiki/Domain-specific_language

here is an example of a DSL for array manipulation,
and an implementation of Floyd Warshall algorithm:

:- op(100, yf, []).
:- op(800, fx, new).
:- op(800, fx, let).
:- op(800, fx, if).

warshall(N, D) :-
    new D[N,N],
    (between(1,N,U), between(1,N,V), let D[U,V] = 999, fail; true),
    (edge(U,V,W), let D[U,V] = W, fail; true),
    (vertex(V), let D[V,V] = 0, fail; true),
    (between(1,N,K),
        between(1,N,I),
           between(1,N,J),
               let H = D[I,K] + D[K,J],
               (if D[I,J] > H -> let D[I,J] = H; true),
               fail; true).

The definition of the DSL needs only one extension
of Prolog, i.e. nb_setarg/3 (SWI-Prolog) respectively
change_arg/3 (Dogelog Player):

new D[N,M] :-
    functor(D, '', N),
    D =.. [_|L],
    new2(L, M).

new2([], _).
new2([X|L], N) :-
    functor(X, '', N),
    new2(L, N).

let V = E :- var(V), !,
    let2(E,V).
let D[R,C] = E :-
    let2(E,V),
    arg(R, D, H),
    nb_setarg(C, H, V).

let2(D[R,C], V) :- !,
    arg(R, D, H),
    arg(C, H, V).
let2(E+F, R) :- !,
    let2(E, V),
    let2(F, W),
    R is V+W.
let2(V, V).

if E > F :-
    let2(E, V),
    let2(F, W),
    V > W.

Idiot Prolog systems like Scryer Prolog or Trealla Prolog
refuse to provide such imperative gadgets, which are quite
useful. If you interpret the DSL, its already bleeing fast,

much faster than a pure implementation:

?- time((between(1,1000,_), graph(G),
    floyd_warshall(4, G, M), fail; true)).
% 3,803,998 inferences, 0.156 CPU in 0.183 seconds (85% CPU, 24345587 Lips)
true.

?- time((between(1,1000,_), warshall(4,D), fail; true)).
% 1,046,998 inferences, 0.062 CPU in 0.062 seconds (100% CPU, 16751968 Lips)
true.

If you compile the DSL, you can again an itch more speed:

/* DSL compiled */
?- time((between(1,1000,_), warshall(4,D), fail; true)).
% 336,998 inferences, 0.000 CPU in 0.020 seconds (0% CPU, Infinite Lips)
true.

Bye

Julio Di Egidio schrieb:
> On 03/01/2025 21:04, Julio Di Egidio wrote:
>> Partial and tentative:
>>
>> ```
>>    Functional = Closures/applications, Reduction/canonicity
>>      /    |
>> Logical  |   = Predicates/queries, Resolution/subsumption
>>      \    |
>>    Imperative = Procedures/invocations, Execution/...
>> ```
>>
>> And there are two views of that triangle: Logical is the top of the 
>> *ideal* such triangle, along the lines of a universe with Prop on top, 
>> which we can reason with; Imperative is the bottom of a *concrete* 
>> such triangle, the bootstrap as well as the final point of application 
>> of any concrete system.
>>
>> And Logical is the constructive (structural) type-theory founding the 
>> Functional, where Functional exists for expressivity and modularity 
>> (what else?), plus can be compiled back/down to machine language...
>>
>> Right?
> 
> BTW, there are deficiencies of standard Prolog that are indeed very 
> annoying, to the point that some invoke for the other way round:
> 
> HANSEI / Re-thinking Prolog
> <https://okmij.org/ftp/kakuritu/logic-programming.html#vs-prolog>
> 
> But, besides that I would not put logic in terms of "guessing", I'd 
> propose we just need a Prolog that doesn't have the self-inflicted 
> quirks: a strengthened resolution with declarative determinism and 
> indexing, and a strengthened semantics, of variables and/vs open terms, 
> with a partial order of terms by subsumption, and unifiability as 
> comparability (a purely structural type system definitionally), i.e. 
> where a variable is the most general term...  Or something like that.
> 
> No?
> 
> -Julio
>