Related
The following is the classic "textbook" vanilla meta-interpreter for prolog.
% simplest meta-interpreter
solve(true) :- !.
solve((A,B)):- !, solve(A), solve(B).
solve(A) :- clause(A,B), solve(B).
The following is simple program which establishes facts two relations which are "positive" and one relation which makes use of negation by failure \+.
% fruit
fruit(apple).
fruit(orange).
fruit(banana).
% colour
yellow(banana).
% Mary likes all fruit
likes(mary, X) :- fruit(X).
% James likes all fruit, as long as it is yellow
likes(james, X) :- fruit(X), yellow(X).
% Sally likes all fruit, except yellow fruit
likes(sally, X) :- fruit(X), \+ (yellow(X)).
The meta-interpeter can handle goals related to the first two relations ?-solve(likes(mary,X)) and ?- solve(likes(james,X)_.
However it fails with a goal related to the third relation ?- solve(likes(sally,X). The swi-prolog reports a stack limit being reached before the program crashes.
Question 1: What is causing the meta-interpreter to fail? Can it be easily adjusted to cope with the \+ negation? Is this related to the sometimes discussed issue of built-ins not being executed by the vanilla meta-interpreter?
Question 2: Where can I read about the need for those cuts in the vanilla meta-interpreter?
Tracing suggests the goal is being grown endlessly:
clause(\+call(call(call(call(yellow(apple))))),_5488)
Exit:clause(\+call(call(call(call(yellow(apple))))),\+call(call(call(call(call(yellow(apple)))))))
Call:solve(\+call(call(call(call(call(yellow(apple)))))))
Call:clause(\+call(call(call(call(call(yellow(apple)))))),_5508)
Exit:clause(\+call(call(call(call(call(yellow(apple)))))),\+call(call(call(call(call(call(yellow(apple))))))))
Call:solve(\+call(call(call(call(call(call(yellow(apple))))))))
Change solve(A) into:
solve(Goal) :-
writeln(Goal),
sleep(1),
clause(Goal, Body),
solve(Body).
... and we see:
?- solve_mi(likes(sally,X)).
likes(sally,_8636)
fruit(_8636)
\+yellow(apple)
\+call(yellow(apple))
\+call(call(yellow(apple)))
\+call(call(call(yellow(apple))))
...
clause/2 determines the body of \+yellow(apple) to be \+call(yellow(apple)), which is not a simplification.
Can use instead:
solve_mi(true) :-
!.
solve_mi((Goal1, Goal2)):-
!,
solve_mi(Goal1),
solve_mi(Goal2).
solve_mi(\+ Goal) :-
!,
\+ solve_mi(Goal).
solve_mi(Goal) :-
clause(Goal, Body),
solve_mi(Body).
Result in swi-prolog:
?- solve_mi(likes(sally,X)).
X = apple ;
X = orange ;
false.
I'm using solve_mi because solve conflicts with e.g. clpBNR, and I'm not using variable names A and B because they convey no meaning.
For understanding the cuts, I'd recommend gtrace, to see the unwanted unification with other goals that would otherwise take place.
All of these predicates are defined in pretty much the same way. The base case is defined for the empty list. For non-empty lists we unify in the head of the clause when a certain predicate holds, but do not unify if that predicate does not hold. These predicates look too similar for me to think it is a coincidence. Is there a name for this, or a defined abstraction?
intersect([],_,[]).
intersect(_,[],[]).
intersect([X|Xs],Ys,[X|Acc]) :-
member(X,Ys),
intersect(Xs,Ys,Acc).
intersect([X|Xs],Ys,Acc) :-
\+ member(X,Ys),
intersect(Xs,Ys,Acc).
without_duplicates([],[]).
without_duplicates([X|Xs],[X|Acc]) :-
\+ member(X,Acc),
without_duplicates(Xs,Acc).
without_duplicates([X|Xs],Acc) :-
member(X,Acc),
without_duplicates(Xs,Acc).
difference([],_,[]).
difference([X|Xs],Ys,[X|Acc]) :-
\+ member(X,Ys),
difference(Xs,Ys,Acc).
difference([X|Xs],Ys,Acc) :-
member(X,Ys),
difference(Xs,Ys,Acc).
delete(_,[],[]).
delete(E,[X|Xs],[X|Ans]) :-
E \= X,
delete(E,Xs,Ans).
delete(E,[X|Xs],Ans) :-
E = X,
delete(E,Xs,Ans).
There is an abstraction for "keep elements in list for which condition holds".
The names are inclide, exclude. There is a library for those in SWI-Prolog that you can use or copy. Your predicates intersect/3, difference/3, and delete/3 would look like this:
:- use_module(library(apply)).
intersect(L1, L2, L) :-
include(member_in(L1), L2, L).
difference(L1, L2, L) :-
exclude(member_in(L2), L1, L).
member_in(List, Member) :-
memberchk(Member, List).
delete(E, L1, L) :-
exclude(=(E), L1, L).
But please take a look at the implementation of include/3 and exclude/3, here:
https://www.swi-prolog.org/pldoc/doc/_SWI_/library/apply.pl?show=src#include/3
Also in SWI-Prolog, in another library, there are versions of those predicates called intersection/3, subtract/3, delete/3:
https://www.swi-prolog.org/pldoc/doc/_SWI_/library/lists.pl?show=src#intersection/3
https://www.swi-prolog.org/pldoc/doc/_SWI_/library/lists.pl?show=src#subtract/3
https://www.swi-prolog.org/pldoc/doc_for?object=delete/3
Those are similar in spirit to your solutions.
Your next predicate, without_duplicates, cannot be re-written like that with include/3 or exclude/3. Your implementation doesn't work, either. Try even something easy, like:
?- without_duplicates([a,b], L).
What happens?
But yeah, it is not the same as the others. To implement it correctly, depending on whether you need the original order or not.
If you don't need to keep the initial order, you can simply sort; this removes duplicates. Like this:
?- sort(List_with_duplicates, No_duplicates).
If you want to keep the original order, you need to pass the accumulated list to the recursive call.
without_duplicates([], []).
without_duplicates([H|T], [H|Result]) :-
without_duplicates_1(T, [H], Result).
without_duplicates_1([], _, []).
without_duplicates_1([H|T], Seen0, Result) :-
( memberchk(H, Seen0)
-> Seen = Seen0 , Result = Result0
; Seen = [H|Seen0], Result = [H|Result0]
),
without_duplicates_1(T, Seen, Result0).
You could get rid of one argument if you use a DCG:
without_duplicates([], []).
without_duplicates([H|T], [H|No_duplicates]) :-
phrase(no_dups(T, [H]), No_duplicates).
no_dups([], _) --> [].
no_dups([H|T], Seen) -->
{ memberchk(H, Seen) },
!,
no_dups(T, Seen).
no_dups([H|T], Seen) -->
[H],
no_dups(T, [H|Seen]).
Well, these are the "while loops" of Prolog on the one hand, and the inductive definitions of mathematical logic on the other hand (See also: Logic Programming, Functional Programming, and Inductive Definitions, Lawrence C. Paulson, Andrew W. Smith, 2001), so it's not surprising to find them multiple times in a program - syntactically similar, with slight deviations.
In this case, you just have a binary decision - whether something is the case or not - and you "branch" (or rather, decide to not fail the body and press on with the selected clause) on that. The "guard" (the test which supplements the head unification), in this case member(X,Ys) or \+ member(X,Ys) is a binary decision (it also is exhaustive, i.e. covers the whole space of possible X)
intersect([X|Xs],Ys,[X|Acc]) :- % if the head could unify with the goal
member(X,Ys), % then additionally check that ("guard")
(...action...). % and then do something
intersect([X|Xs],Ys,Acc) :- % if the head could unify with the goal
\+ member(X,Ys), % then additionally check that ("guard")
(...action...). % and then do something
Other applications may need the equivalent of a multiple-decision switch statement here, and so N>2 clauses may have to be written instead of 2.
foo(X) :-
member(X,Set1),
(...action...).
foo(X) :-
member(X,Set2),
(...action...).
foo(X) :-
member(X,Set3),
(...action...).
% inefficient pseudocode for the case where Set1, Set2, Set3
% do not cover the whole range of X. Such a predicate may or
% may not be necessary; the default behaviour would be "failure"
% of foo/1 if this clause does not exist:
foo(X) :-
\+ (member(X,Set1);member(X,Set2);member(X,Set3)),
(...action...).
Note:
Use memberchk/2 (which fails or succeeds-once) instead of member/2 (which fails or succeeds-and-then-tries-to-succeed-again-for-the-rest-of-the-set) to make the program deterministic in its decision whether member(X,L).
Similarly, "cut" after the clause guard to tell Prolog that if a guard of one clause succeeds, there is no point in trying the other clauses because they will all turn out false: member(X,Ys),!,...
Finally, use term comparison == and \== instead of unification = or unification failure \= for delete/3.
I'm writing prolog code that finds a certain number; a number is the right number if it's between 0 and 9 and not present in a given list. To do this I wrote a predicate number/3 that has the possible numbers as the first argument, the list in which the Rightnumber cannot be present and the mystery RightNumber as third argument:
number([XH|XT], [H|T], RightNumber):-
member(XH, [H|T]), !,
number(XT, [H|T], RightNumber).
number([XH|_], [H|T], XH):-
\+ member(XH, [H|T]).
so this code basically says that if the Head of the possible numbers list is already a member of the second list, to cut of the head and continue in recursion with the tail.
If the element is not present in the second list, the second clause triggers and tells prolog that that number is the RightNumber. It's okay that it only gives the first number that is possible, that's how I want to use it.
This code works in theory, but I was wondering if there's a better way to write it down? I'm using this predicate in another predicate later on in my code and it doesn't work as part of that. I think it's only reading the first clause, not the second and fails as a result.
Does anybody have an idea that might improve my code?
sample queries:
?- number([0,1,2,3,4,5,6,7,8,9], [1,2], X).
X = 3
?- number([0,1,2,3,4,5,6,7,8,9], [1,2,3,4,5,6,7,8,0], X).
X = 9
First, the code does not work. Consider:
?- number(Xs, Ys, N).
nontermination
This is obviously bad: For this so-called most general query, we expect to obtain answers, but Prolog does not give us any answer with this program!
So, I first suggest you eliminate all impurities from your program, and focus on a clean declarative description of what you want.
I give you a start:
good_number(N, Ls) :-
N in 0..9,
maplist(#\=(N), Ls).
This states that the relation is true if N is between 0 and 9, and N is different from any integer in Ls. See clpfd for more information about CLP(FD) constraints.
Importantly, this works in all directions. For example:
?- good_number(4, [1,2,3]).
true.
?- good_number(11, [1,2,3]).
false.
?- good_number(N, [1,2,3]).
N in 0\/4..9.
And also in the most general case:
?- good_number(N, Ls).
Ls = [],
N in 0..9 ;
Ls = [_2540],
N in 0..9,
N#\=_2540 ;
Ls = [_2750, _2756],
N in 0..9,
N#\=_2756,
N#\=_2750 .
This, with only two lines of code, we have implemented a very general relation.
Also see logical-purity for more information.
First of all, your predicate does not work, nor does it check all the required constraints (between 0 and 9 for instance).
Several things:
you unpack the second list [H|T], but you re-pack it when you call member(XH, [H|T]); instead you can use a list L (this however slightly alters the semantics of the predicate, but is more accurate towards the description);
you check twice member/2ship;
you do not check whether the value is a number between 0 and 9 (and an integer anyway).
A better approach is to construct a simple clause:
number(Ns, L, Number) :-
member(Number, Ns),
integer(Number),
0 =< Number,
Number =< 9,
\+ member(Number, L).
A problem that remains is that Number can be a variable. In that case integer(Number) will fail. In logic we would however expect that Prolog unifies it with a number. We can achieve this by using the between/3 predicate:
number(Ns, L, Number) :-
member(Number, Ns),
between(0, 9, Number),
\+ member(Number, L).
We can also use the Constraint Logic Programming over Finite Domains library and use the in/2 predicate:
:- use_module(library(clpfd)).
number(Ns, L, Number) :-
member(Number, Ns),
Number in 0..9,
\+ member(Number, L).
There are still other things that can go wrong. For instance we check non-membership with \+ member(Number, L). but in case L is not grounded, this will fail, instead of suggesting lists where none of the elements is equal to Number, we can use the meta-predicate maplist to construct lists and then call a predicate over every element. The predicate we want to call over every element is that that element is not equal to Number, so we can use:
:- use_module(library(clpfd)).
number(Ns, L, Number) :-
member(Number, Ns),
Number in 0..9,
maplist(#\=(Number), L).
Let's say there is a simple database of people in Prolog
person(john).
person(mary).
person(john).
person(susan).
I need to match the entires exactly once:
john-mary, john-john, john-susan, mary-john, mary-susan, john-susan
I tried coming up with something like this:
match:- person(X),!,person(Y), write(X),write(-), write(Y),nl.
run:- person(X), match(X), fail.
But it's matching many times, and matches a person to him/herself, which shouldn't be.
Basically, what I need is to iterate over all Xs and make Prolog to look strictly "below" for Ys.
A quick solution would be to number your people:
person(1, john).
person(2, mary).
person(3, john).
person(4, susan).
Then you could match people like this:
match(X-Y) :-
person(I, X), person(J, Y), I < J.
Since you have two john entries, I'm not sure any other solution is going to work. Normally you could fake an ordering using #>/2 but that would require your atoms to be unique, and since they aren't, it would prevent the john-john solution.
Edit: Since we're willing to use findall/3 to materialize the database of people, we can treat this as a list problem and find a functional solution. Let's get all the combinations in a list:
combinations([X|Rest], X, Y) :- member(Y, Rest).
combinations([_|Rest], X, Y) :- combinations(Rest, X, Y).
With this predicate in hand, we can find the solution:
combined_folks(People) :-
findall(P, person(P), Persons),
findall(X-Y, combinations(Persons, X, Y), People).
?- combined_folks(X).
X = [john-mary, john-john, john-susan, mary-john, mary-susan, john-susan].
That actually turned out to be pretty clean!
person(john).
person(mary).
person(john).
person(susan).
match :- findall(P,person(P),People), match_all(People).
match_all([_]) :- !.
match_all([P|People]) :- match_2(P,People), match_all(People).
match_2(_,[]) :- !.
match_2(P1,[P2|People]) :- format('~a-~a~n',[P1,P2]), match_2(P1,People).
?- match.
Is it possible to have a "variable arity predicate" in Prolog?
I mean something like this:
my_predicate( [a,b,c], [a,c], [a], [a,b,c,d], N, RESULT)
with the number of initial lists unknown at the beginning?
Using the univ operator ( =.. ) it would be possible to unify it with a list of terms and traversing it like every other list. But how to write the goal?
my_predicate(??) =.. [??]
I really don't know if this is even possible..
you can define predicates with different arities that have the same name but they will be different predicates.
foo(1).
foo(2,1).
?-foo(2).
false
my suggestion is to change the encoding; instead of a number of initial lists, have a list of initial lists.
the other solution would be to write predicates for all the possible numbers of arguments (or dynamically generate them).
As #thanosQR suggests, it probably is the best to change your representation to some list.
There are, however - very seldom but nevertheless - situations where you want to define a predicate for many different arities. In this very rare case, you can define such a predicate manually. That is, for each arity manually. Of course, you will only define several cases. As an example, see library(lambda).
You always can go up to one level :) Many years ago I saw implementation of ANSI prolog interpter in Turbo Prolog. Idea was very simple, enclose all user-space facts and rules in single fact backed by assert/retract-like operations.
Consider enclosing all your targets in another compose:
target(my_predicate( [a,b,c], [a,c], [a], [a,b,c,d], N, RESULT)) :- RESULT=[a], N=1.
target(H) :- H =.. [my_predicate|_].
target(using_my_predicate(X, Y)) :- target(my_predicate(X,1,Y)).
Some prologs (at least YAP) have directives to declare handlers for unknown targets:
:- module(sumtest).
target(sum(0)).
target(H) :-
H =.. [sum, S, X|XS],
H1 =.. [sum, S1|XS],
H1,
S is (S1+X).
target(sumtest:G):- target(G). % HACK: strip-off module
:- unknown(_, target(_)).
test:-
sum(X,1), write(X), nl,
sum(Y,2,3), write(Y), nl,
sum(Z,3,4,2), write(Z), nl,
target(sum(X1,1)), write(X1), nl,
target(sum(Y1,2,3)), write(Y1), nl,
target(sum(Z1,3,4,2)), write(Z1), nl.
:- test, halt.
% % yap -l sumtest.pl
% YAP 6.2.0 (amd64): Thu Oct 21 10:31:27 EEST 2010
% MYDDAS version MYDDAS-0.9.1
% 1
% 5
% 9
% 1
% 5
% 9
% % YAP execution halted