Re: [eclipse-users] Matching

Lutz, Charles D wrote:
> To be honest, the outputs of both of these variants are not totally 
> clear to me. First off,
> 
> what does it “mean” to have “f(a, X)” inside of the head clause? Apart 
> from saying p1
> 
> or p2 is a clause of arity 1, whose argument is another clause of arity 
> 2, what is the
> 
> “practical” meaning of it?

We have to be careful with the terminology:

- we have here two _predicates_ (p1/1 and p2/1),
   i.e. things that can be true of false.

- each predicate is defined by (one or more) _clauses_.
   Here, we only have one clause each for p1/1 and p2/1.

Since there are no clauses for f/2, there is no predicate f/2,
i.e. it doesn't make sense to ask whether f/2 is true or false.

f(a,X) is therefore simply a data structure, more precisely a
compound term with two subterms.


Now, a clause like

p1(f(a,X)) :- ...

means that it provides a definition for the truth of p1/1, but it
applies only to the case that the argument of p1/1 is a data structure
of the form f(a,X).

Incidentally, this is equivalent to having written

p1(Arg) :- Arg = f(a,X), ...


To understand what happens when you ask the query (your example 24)

?- p1(f(a,b)).

in the presence of the definition:

p1(Arg) :- Arg = f(a,X), writeln(X).

you can simply "macro-expand" the definition, by substituting in your
query the left-hand side of the defining clause with the right hand
side, and filling in the actual value for Arg.  This gives

?- f(a,b) = f(a,X), writeln(X).

and this should explain why it writes "b".


Now look at your query 23, whose expansion yields

?- f(A,b) = f(a,X) , writeln(X).

This still writes "b", but since unification (=) works both ways,
it also binds the variable A to a.


The "matching clauses" work the same, except that they use a one-way
variant of unification, which only allows variables on the right hand
side to be bound (and fails otherwise).


Apart from the matching clauses (which are ECLiPSe-specific), this is
standard Prolog mechanism which you can find in quite compact form in
e.g. the text book like Bratko, or in the ECLiPSe book by Apt/Wallace,
or have a look at some of the online resources at
www.eclipse-clp.org/links.html

It would be great to beef up the ECLiPSe tutorial and make it more self
contained.  If someone feels inspired, the sources are available at
SourceForge!


Best regards,
Joachim