09.07.2018 11:50, Dušan Kolář wrote:

Hmm, but CLP has completely different way of evaluation and program construction than Prolog has. Moreover, stating it is a Prolog is not so much true then. That's the point as I can see it.

 

I got you. And this was my point, "declarative" for me, means some declarations/restrictions about answer/answers. No exact algorithm: you can even don't know how this is evaluated/executed. You only find some relations and declare them. And set of solvers tries to find solutions. So, modern Prolog/Logic systems are Prolog as DSL and set of solvers, right. For example, to find factorial as well as it's argument, you need to implement CLP library in Haskell explicitly (if you want to use CLP way). This is the Haskell tradition. Sure, you can do it in Python, in C (sure, such libraries already exist). But it's not declarative tradition: predicates calculus is based on backtracking and variables unification. Haskell, like Scheme is based on lambda calculus. So, evaluation is totally different.

But more interesting here (regarding to Bool subj)  is that Haskell can return to you 120 as factorial result, but Prolog will answer you (if it finds several results) "do you need next one?" and result in Prolog is "Yes.". This is important. Which can be imagine in terms of Haskell like you have not top-level `IO ()` but some `LogicMonad ()`. And all evaluations in Haskell will lead to Boolean (OK, truth/lie/yes/no) result. Evaluated (integers/floats/etc) result must be returned as output variables (your factorial is: `factorial(+N, -F)` i.e. it's one-directional only, so all functions in such Haskell also can return results in some reference/output var and evaluated to Bool). I can imagine this. Also if result is "No.", output variables unifications (bindings - in Haskell terms) will be dropped out. Unfortunately I dont know Curry and absolutely can not say: is it implemented in such way. But I'm sure that it's possible :)

Joachim pointed out "cut" operation which leads to more effective code, but eliminates pure logic constructs. There are "green cuts" (does not change semantic) and "red cuts" and this shows that practical real world languages mostly are not pure functional/pure logical.

But there is another point: it makes sense to have pure logical application which can be driven from C# or Java application (there are such puzzles and other games, on Android, Windows, tables/plane schedulers, expert systems, etc), but I can not imagine why I'll call for example, Haskell from the Java, because their area mostly are intersected in contrast with Java/Prolog, for example. This is some marker too, about which languages have similar "nature".

I have already agreed that are possible another classifications coordinates as well as it's difficult to classify some languages because they are totally hybrid (I remember yet another such: Racket, Oz/Mozart).

PS. I'm application developer only, not researcher, but I'm absolutely sure: there are people in the list which are totally in the subject, even may be some of them are devs of Curry or Mercury for example, and maybe they are laughing at me and my naive intuitions now :)


And it has nothing to do with Haskell extensions. In Haskell, the principle remains the same, though.

 

 

Dušan

 

On pondělí 9. července 2018 10:22:15 CEST you wrote:

> Hello, Dušan!

>

> > This is not Prolog, it is a Prolog with CLP extension over integers.

>

> as well as Haskell has a lot of extensions. I mentioned CLP as solver

> but modern Prologs is not ISO only and has a lot of solvers (CHR,

> different CLP, FuzzyLogic, Tabulating, etc). Actually CLP(x) in included

> in most modern implementations.

>

> 09.07.2018 10:58, Dušan Kolář wrote:

> > Hello PY/Paul,

> >

> > You are using misleading arguments!

> >

> > This is not Prolog, it is a Prolog with CLP extension over integers.

> >

> > > factorial(0, 1).

> > >

> > > factorial(N, F) :-

> > >

> > > N #> 0,

> > >

> > > N1 #= N - 1,

> > >

> > > F #= N * F1,

> > >

> > > factorial(N1, F1).

> > >

> > >

> > >

> > > I can call it: factorial(5, X) => X=120.

> > >

> > > But also: factorial(X, 120) => X=5.

> >

> > Pure Prolog is:

> >

> > factorial(0, 1) :- !.

> >

> > factorial(N, F) :-

> >

> > NN is N - 1,

> >

> > factorial(NN, FF),

> >

> > F is FF * N.

> >

> > which is a function, not a relation, as you write.

> >

> > Dušan Kolář