Prolog removing unique elements only - list

I want to return a list that removes all unique elements for example
remUniqueVals([1,1,2,2,3,4,4,5,6,6,6],Q).
Q = [1,1,2,2,4,4,6,6,6].
My problem is that currently I have code that returns
remUniqueVals([1,1,2,2,3,4,4,5,6,6,6],Q).
Q = [1, 2, 4, 6, 6].
So that only the first instance of these non-unique values are returned.
Here is my code:
remUniqueVals([], []).
remUniqueVals([Q1|RestQ],[Q1|Xs]) :-
member(Q1,RestQ),
remUniqueVals(RestQ,Xs).
remUniqueVals([Q1|RestQ],Xs) :-
remove(Q1,[Q1|RestQ], NewQ),
remUniqueVals(NewQ,Xs).
I can see that member(Q1,RestQ) fails when it checks 1,2,4 the second time because they are now no longer in the list and so removes them. I'd like some helping solving this problem, my thoughts are to check member(Q1, PreviousQ), where this is the elements already in the final Q. Not sure how to go about implementing that though any help would be appreciated.
Update:
Ok so thanks for the suggestions I ended up going with this in the end:
remUniqueVals(_,[], []).
remUniqueVals(_,[Q1|RestQ],[Q1|Xs]) :-
member(Q1,RestQ),
remUniqueVals(Q1,RestQ,Xs).
remUniqueVals(PrevQ,[Q1|RestQ],[Q1|Xs]) :-
Q1 = PrevQ,
remUniqueVals(PrevQ,RestQ,Xs).
remUniqueVals(PrevQ,[_|RestQ],Xs) :-
remUniqueVals(PrevQ,RestQ,Xs).
remUniqueVals(0,[4,1,1,3,2,2,5,5],Q).
Q = [1, 1, 2, 2, 5, 5].
remUniqueVals(0, [A,B,C], [1,1]).
A = 1,
B = 1,
C = 1.

Prolog rules are read independently of each other, so you need one rule for the case where the element is unique and one where it is not. Provided the order of the elements is not relevant, you might use:
?- remUniqueVals([A,B,C], [1,1]).
A = 1, B = 1, dif(1,C)
; A = 1, C = 1, dif(1,B)
; B = 1, C = 1, dif(A,1)
; false.
?- remUniqueVals([1,1,2,2,3,4,4,5,6,6,6],Q).
Q = [1,1,2,2,4,4,6,6,6]
; false.
remUniqueVals([], []).
remUniqueVals([Q1|RestQ],[Q1|Xs0]) :-
memberd(Q1, RestQ),
phrase(delall(Q1, RestQ, NewQ), Xs0, Xs),
remUniqueVals(NewQ, Xs).
remUniqueVals([Q1|RestQ],Xs) :-
maplist(dif(Q1), RestQ),
remUniqueVals(RestQ,Xs).
memberd(X, [X|_Xs]).
memberd(X, [Y|Xs]) :-
dif(X,Y),
memberd(X, Xs).
delall(_X, [], []) --> [].
delall(X, [X|Xs], Ys) -->
[X],
delall(X, Xs, Ys).
delall(X, [Y|Xs], [Y|Ys]) -->
{dif(X,Y)},
delall(X, Xs, Ys).
Here is an alternate definition for memberd/2 which might be more efficient using if_/3:
memberd(E, [X|Xs]) :-
if_(E = X, true, memberd(E, Xs) ).

This is similar to the original solution but it collects the non-unique values in an auxiliary list and checks it to avoid removing the last one from the original:
remove_uniq_vals(L, R) :-
remove_uniq_vals(L, [], R).
remove_uniq_vals([], _, []).
remove_uniq_vals([X|T], A, R) :-
( member(X, A)
-> R = [X|T1], A1 = A
; member(X, T)
-> R = [X|T1], A1 = [X|A]
; R = T1, A1 = A
),
remove_uniq_vals(T, A1, T1).
Testing...
| ?- remove_uniq_vals([1,2,3,1,2,3,1,2,3,4,3], Q).
Q = [1,2,3,1,2,3,1,2,3,3]
(1 ms) yes
| ?- remove_uniq_vals([1,1,2,2,3,4,4,5,6,6,6], Q).
Q = [1,1,2,2,4,4,6,6,6]
yes
So the predicate works great if the first argument is an input, and it maintains the original order of the remaining elements in the list.
However, this predicate is not completely relational in that it will fail a case in which the first argument is an uninstantiated list of a known number of elements and the second argument is a list of a different fixed number of elements. So something like this will work:
| ?- remove_uniq_vals([A,B,C], L).
B = A
C = A
L = [A,A,A]
(1 ms) yes
But something like the following fails:
| ?- remove_uniq_vals([A,B,C], [1,1]).
no

This is another pure, relational solution inspired by #CapelliC's solution. This one now retains the order of the duplicates. What is interesting to see is how the implicit quantification happening in #CapelliC's solution now has to be done explicitly.
The biggest advantage of having a pure, relational definition is that noes are noes. And ayes are ayes. That is: You do not have to worry whether or not the answer you get happens to be correct or not. It is correct (or incorrect — but it is not partially correct). Non-relational solutions can often be cleansed by producing instantiation_error in case the method fails. But as you can verify yourself, both have "forgotten" such tests thereby preparing a nice habitat for bugs. A safe test for those other solutions would have been ground(Xs) or ground(Xs), acyclic_term(Xs) but much too often this is considered too restricted.
remUniqueVals2(Xs, Ys) :-
tfilter(list_withduplicate_truth(Xs),Xs,Ys).
list_withduplicate_truth(L, E, Truth) :-
phrase(
( all(dif(E)),
( {Truth = false}
| [E],
all(dif(E)),
( {Truth = false}
| {Truth = true},
[E],
...
)
)
), L).
all(_) --> [].
all(P_1) -->
[E],
{call(P_1,E)},
all(P_1).
... --> [] | [_], ... .
tfilter( _, [], []).
tfilter(TFilter_2, [E|Es], Fs0) :-
call(TFilter_2,E,Truth),
( Truth = false,
Fs0 = Fs
; Truth = true,
Fs0 = [E|Fs]
),
tfilter(TFilter_2, Es, Fs).
Another, more compact way using if_/3
tfilter( _, [], []).
tfilter(TFilter_2, [E|Es], Fs0) :-
if_(call(TFilter_2,E), Fs0 = [E|Fs], Fs0 = Fs ),
tfilter(TFilter_2, Es, Fs).

Preserve logical-purity! Based on if_/3, (=)/3, and meta-predicate tpartition/4 we define:
remUniqueValues([], []).
remUniqueValues([X|Xs1], Ys1) :-
tpartition(=(X), Xs1, Eqs, Xs0),
if_(Eqs = [],
Ys1 = Ys0,
append([X|Eqs], Ys0, Ys1)),
remUniqueValues(Xs0, Ys0).
Let's see it in action!
?- remUniqueValues([A,B,C], [1,1]).
A=1 , B=1 , dif(C,1)
; A=1 , dif(B,1), C=1
; dif(A,1), B=1 , C=1
; false.
?- remUniqueValues([1,1,2,2,3,4,4,5,6,6,6], Vs).
Vs = [1,1,2,2,4,4,6,6,6]. % succeeds deterministically

This is a purified version of #mbratch's solution. It uses a reïfied version of member/2 which is free of redundant answers like for member(X,[a,a]).
memberd_truth_dcg(X, Xs, Truth) :-
phrase(( all(dif(X)), ( [X], {Truth = true}, ... | {Truth = false} ) ), Xs).
A slightly generalized version which only requires to have a list prefix, but not a list:
memberd_truth(_X, [], false).
memberd_truth(X, [X|_], true).
memberd_truth(X, [Y|Ys], Truth) :-
dif(X,Y),
memberd_truth(X, Ys, Truth).
The variables are named in the same manner as in #mbratch's solution:
remove_uniq_valsBR(L, R) :-
remove_uniq_valsBR(L, [], R).
remove_uniq_valsBR([], _, []).
remove_uniq_valsBR([X|T], A, R) :-
memberd_truth(X, A, MemT1),
( MemT1 = true,
R = [X|T1], A1 = A
; MemT1 = false,
memberd_truth(X, T, MemT2),
( MemT2 = true,
R = [X|T1], A1 = [X|A]
; MemT2 = false,
R = T1, A1 = A
)
),
remove_uniq_valsBR(T, A1, T1).
More compactly using if/3:
remove_uniq_valsBR([], _, []).
remove_uniq_valsBR([X|T], A, R) :-
if_( memberd_truth(X, A),
( R = [X|T1], A1 = A ),
if_( memberd_truth(X, T),
( R = [X|T1], A1 = [X|A] ),
( R = T1, A1 = A ) ) )
),
remove_uniq_valsBR(T, A1, T1).
What I do not like is the many redundant dif/2 constraints. I hoped this version would have less of them:
?- length(L,_),remove_uniq_valsBR(L,L).
L = []
; L = [_A,_A]
; L = [_A,_A,_A]
; L = [_A,_A,_A,_A]
; L = [_A,_A,_B,_B], dif(_B,_A)
; L = [_A,_B,_A,_B],
dif(_A,_B), dif(_B,_A), dif(_B,_A), dif(_A,_B)
; ... .
Of course it is possible to check whether or not a dif/2 is already present, but I'd prefer a version where there are fewer dif/2 goals posted right from the beginning.

a solution based on 3 builtins:
remUniqueVals(Es, NUs) :-
findall(E, (select(E, Es, R), memberchk(E, R)), NUs).
can be read as
find all elements that still appear in list after have been selected

Related

Remove brackets from a list in Prolog [duplicate]

I am doing some easy exercises to get a feel for the language.
is_list([]).
is_list([_|_]).
my_flatten([],[]).
my_flatten([X|Xs],RR) :-
my_flatten(Xs,R),
(is_list(X), !, append(X,R,RR); RR = [X | R]).
Here is a version using cut, for a predicate that flattens a list one level.
my_flatten([],[]).
my_flatten([X|Xs],RR) :-
my_flatten(Xs,R),
if_(is_list(X), append(X,R,RR), RR = [X | R]).
Here is how I want to write it, but it does not work. Neither does is_list(X) = true as the if_ condition. How am I intended to use if_ here?
(Sorry, I somewhat skipped this)
Please refer to P07. It clearly states that it flattens out [a, [b, [c, d], e]], but you and #Willem produce:
?- my_flatten([a, [b, [c, d], e]], X).
X = [a,b,[c,d],e]. % not flattened!
And the solution given there succeeds for
?- my_flatten(non_list, X).
X = [non_list]. % unexpected, nothing to flatten
Your definition of is_list/1 succeeds for is_list([a|non_list]). Commonly, we want this to fail.
What you need is a safe predicate to test for lists. So let's concentrate on that first:
What is wrong with is_list/1 and if-then-else? It is as non-monotonic, as many other impure type testing predicates.
?- Xs = [], is_list([a|Xs]).
Xs = [].
?- is_list([a|Xs]). % generalization, Xs = [] removed
false. % ?!? unexpected
While the original query succeeds correctly, a generalization of it unexpectedly fails. In the monotonic part of Prolog, we expect that a generalization will succeed (or loop, produce an error, use up all resources, but never ever fail).
You have now two options to improve upon this highly undesirable situation:
Stay safe with safe inferences, _si!
Just take the definition of list_si/1 in place of is_list/1. In problematic situations, your program will now abort with an instantiation error, meaning "well sorry, I don't know how to answer this query". Be happy for that response! You are saved from being misled by incorrect answers.
In other words: There is nothing wrong with ( If_0 -> Then_0 ; Else_0 ), as long as the If_0 handles the situation of insufficient instantiations correctly (and does not refer to a user defined program since otherwise you will be again in non-monotonic behavior).
Here is such a definition:
my_flatten(Es, Fs) :-
list_si(Es),
phrase(flattenl(Es), Fs).
flattenl([]) --> [].
flattenl([E|Es]) -->
( {list_si(E)} -> flattenl(E) ; [E] ),
flattenl(Es).
?- my_flatten([a, [b, [c, d], e]], X).
X = [a,b,c,d,e].
So ( If_0 -> Then_0 ; Else_0 ) has two weaknesses: The condition If_0 might be sensible to insufficient instantiations, and the Else_0 may be the source of non-monotonicity. But otherwise it works. So why do we want more than that?
In many more general situations this definition will now bark back: "Instantiation error"! While not incorrect, this still can be improved. This exercise is not the ideal example for this, but we will give it a try.
Use a reified condition
In order to use if_/3 you need a reified condition, that is, a definition that carries it's truth value as an explicit extra argument. Let's call it list_t/2.
?- list_t([a,b,c], T).
T = true.
?- list_t([a,b,c|non_list], T).
T = false.
?- list_t(Any, T).
Any = [],
T = true
; T = false,
dif(Any,[]),
when(nonvar(Any),Any\=[_|_])
; Any = [_],
T = true
; Any = [_|_Any1],
T = false,
dif(_Any1,[]),
when(nonvar(_Any1),_Any1\=[_|_])
; ... .
So list_t can also be used to enumerate all true and false situations. Let's go through them:
T = true, Any = [] that's the empty list
T = false, dif(Any, []), Any is not [_|_] note how this inequality uses when/2
T = true, Any = [_] that's all lists with one element
T = true, Any = [_|_Any1] ... meaning: we start with an element, but then no list
list_t(Es, T) :-
if_( Es = []
, T = true
, if_(nocons_t(Es), T = false, ( Es = [_|Fs], list_t(Fs, T) ) )
).
nocons_t(NC, true) :-
when(nonvar(NC), NC \= [_|_]).
nocons_t([_|_], false).
So finally, the reified definition:
:- meta_predicate( if_(1, 2, 2, ?,?) ).
my_flatten(Es, Fs) :-
phrase(flattenl(Es), Fs).
flattenl([]) --> [].
flattenl([E|Es]) -->
if_(list_t(E), flattenl(E), [E] ),
flattenl(Es).
if_(C_1, Then__0, Else__0, Xs0,Xs) :-
if_(C_1, phrase(Then__0, Xs0,Xs), phrase(Else__0, Xs0,Xs) ).
?- my_flatten([a|_], [e|_]).
false.
?- my_flatten([e|_], [e|_]).
true
; true
; true
; ... .
?- my_flatten([a|Xs], [a]).
Xs = []
; Xs = [[]]
; Xs = [[],[]]
; ... .
?- my_flatten([X,a], [a]).
X = []
; X = [[]]
; X = [[[]]]
; X = [[[[]]]]
; ... .
?- my_flatten(Xs, [a]).
loops. % at least it does not fail
In Prolog, the equivalen of an if … then … else … in other languages is:
(condition -> if-true; if-false)
With condition, if-true and if-false items you need to fill in.
So in this specific case, you can implement this with:
my_flatten([],[]).
my_flatten([X|Xs],RR) :-
my_flatten(Xs,R),
( is_list(X)
-> append(X,R,RR)
; RR = [X | R] ).
or we can flatten recursively with:
my_flatten([],[]).
my_flatten([X|Xs],RR) :-
my_flatten(Xs,R),
( flatten(X, XF)
-> append(XF,R,RR)
; RR = [X | R] ).
Your if_/3 predicate is used for reified predicates.
This worked for me:
myflat([], []).
myflat([H|T], L) :-
myflat(H, L1),
myflat(T, L2),
append(L1, L2, L).
myflat(L, [L]).

How to implement a not_all_equal/1 predicate

How would one implement a not_all_equal/1 predicate, which succeeds if the given list contains at least 2 different elements and fails otherwise?
Here is my attempt (a not very pure one):
not_all_equal(L) :-
( member(H1, L), member(H2, L), H1 \= H2 -> true
; list_to_set(L, S),
not_all_equal_(S)
).
not_all_equal_([H|T]) :-
( member(H1, T), dif(H, H1)
; not_all_equal_(T)
).
This however does not always have the best behaviour:
?- not_all_equal([A,B,C]), A = a, B = b.
A = a,
B = b ;
A = a,
B = b,
dif(a, C) ;
A = a,
B = b,
dif(b, C) ;
false.
In this example, only the first answer should come out, the two other ones are superfluous.
Here is a partial implementation using library(reif) for SICStus|SWI. It's certainly correct, as it produces an error when it is unable to proceed. But it lacks the generality we'd like to have.
not_all_equalp([A,B]) :-
dif(A,B).
not_all_equalp([A,B,C]) :-
if_(( dif(A,B) ; dif(A,C) ; dif(B,C) ), true, false ).
not_all_equalp([A,B,C,D]) :-
if_(( dif(A,B) ; dif(A,C) ; dif(A,D) ; dif(B,C) ; dif(B,D) ), true, false ).
not_all_equalp([_,_,_,_,_|_]) :-
throw(error(representation_error(reified_disjunction),'C\'est trop !')).
?- not_all_equalp(L).
L = [_A,_B], dif(_A,_B)
; L = [_A,_A,_B], dif(_A,_B)
; L = [_A,_B,_C], dif(_A,_B)
; L = [_A,_A,_A,_B], dif(_A,_B)
; L = [_A,_A,_B,_C], dif(_A,_B)
; L = [_A,_B,_C,_D], dif(_A,_B)
; error(representation_error(reified_disjunction),'C\'est trop !').
?- not_all_equalp([A,B,C]), A = a, B = b.
A = a, B = b
; false.
Edit: Now I realize that I do not need to add that many dif/2 goals at all! It suffices that one variable is different to the first one! No need for mutual exclusivity! I still feel a bit insecure to remove the dif(B,C) goals ...
not_all_equalp([A,B]) :-
dif(A,B).
not_all_equalp([A,B,C]) :-
if_(( dif(A,B) ; dif(A,C) ), true, false ).
not_all_equalp([A,B,C,D]) :-
if_(( dif(A,B) ; dif(A,C) ; dif(A,D) ), true, false ).
not_all_equalp([_,_,_,_,_|_]) :-
throw(error(representation_error(reified_disjunction),'C\'est trop !')).
The answers are exactly the same... what is happening here, me thinks. Is this version weaker, that is less consistent?
Here's a straightforward way you can do it and preserve logical-purity!
not_all_equal([E|Es]) :-
some_dif(Es, E).
some_dif([X|Xs], E) :-
( dif(X, E)
; X = E, some_dif(Xs, E)
).
Here are some sample queries using SWI-Prolog 7.7.2.
First, the most general query:
?- not_all_equal(Es).
dif(_A,_B), Es = [_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_A,_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_A,_A,_A,_B|_C]
...
Next, the query the OP gave in the question:
?- not_all_equal([A,B,C]), A=a, B=b.
A = a, B = b
; false. % <- the toplevel hints at non-determinism
Last, let's put the subgoal A=a, B=b upfront:
?- A=a, B=b, not_all_equal([A,B,C]).
A = a, B = b
; false. % <- (non-deterministic, like above)
Good, but ideally the last query should have succeeded deterministically!
Enter library(reif)
First argument indexing
takes the principal functor of the first predicate argument (plus a few simple built-in tests) into account to improve the determinism of sufficiently instantiated goals.
This, by itself, does not cover dif/2 satisfactorily.
What can we do? Work with
reified term equality/inequality—effectively indexing dif/2!
some_dif([X|Xs], E) :- % some_dif([X|Xs], E) :-
if_(dif(X,E), true, % ( dif(X,E), true
(X = E, some_dif(Xs,E)) % ; X = E, some_dif(Xs,E)
). % ).
Notice the similarities of the new and the old implementation!
Above, the goal X = E is redundant on the left-hand side. Let's remove it!
some_dif([X|Xs], E) :-
if_(dif(X,E), true, some_dif(Xs,E)).
Sweet! But, alas, we're not quite done (yet)!
?- not_all_equal(Xs).
DOES NOT TERMINATE
What's going on?
It turns out that the implementation of dif/3 prevents us from getting a nice answer sequence for the most general query. To do so—without using additional goals forcing fair enumeration—we need a tweaked implementation of dif/3, which I call diffirst/3:
diffirst(X, Y, T) :-
( X == Y -> T = false
; X \= Y -> T = true
; T = true, dif(X, Y)
; T = false, X = Y
).
Let's use diffirst/3 instead of dif/3 in the definition of predicate some_dif/2:
some_dif([X|Xs], E) :-
if_(diffirst(X,E), true, some_dif(Xs,E)).
So, at long last, here are above queries with the new some_dif/2:
?- not_all_equal(Es). % query #1
dif(_A,_B), Es = [_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_B|_C]
; dif(_A,_B), Es = [_A,_A,_A,_B|_C]
...
?- not_all_equal([A,B,C]), A=a, B=b. % query #2
A = a, B = b
; false.
?- A=a, B=b, not_all_equal([A,B,C]). % query #3
A = a, B = b.
Query #1 does not terminate, but has the same nice compact answer sequence. Good!
Query #2 is still non-determinstic. Okay. To me this is as good as it gets.
Query #3 has become deterministic: Better now!
The bottom line:
Use library(reif) to tame excess non-determinism while preserving logical purity!
diffirst/3 should find its way into library(reif) :)
EDIT: more general using a meta-predicate (suggested by a comment; thx!)
Let's generalize some_dif/2 like so:
:- meta_predicate some(2,?).
some(P_2, [X|Xs]) :-
if_(call(P_2,X), true, some(P_2,Xs)).
some/2 can be used with reified predicates other than diffirst/3.
Here an update to not_all_equal/1 which now uses some/2 instead of some_dif/2:
not_all_equal([X|Xs]) :-
some(diffirst(X), Xs).
Above sample queries still give the same answers, so I won't show these here.

Remove duplicates from a list while keeping the rightmost occurrences

I am trying to remove duplicates from a list while keeping the rightmost occurrences. E.g.: [1,2,3,1,2] is transformed in [3,1,2]
It's one of my first tries in Prolog and I don't understand what am I doing wrong. It always returns false. This is my code:
%nrap(L:list,E:element,S:integer)
%L - the initial list, list of integers
%E - the element, integer
%S - the result, nrap of E in L, S integer
%flow model: (i,i,o),(i,i,i)
nrap([],_,0).
nrap([H|T],E,S):-
H=E,
nrap(T,E,S1),
S is S1+1.
nrap([H|T],E,S):-
H\=E,
nrap(T,E,S).
%transform(L:list,L2:list,R:list)
%L - the initial list, list of integers
%L2 - copy of the initial list
%R - the resulted list, without duplicates, list of integers
%flow model: (i,i,o),(i,i,i)
transform([],[],[]).
transform([H|T],L2,[H|R]):-
nrap(L2,H,S),
S=1,
transform(T,L2,R).
transform([H|T],L2,R):-
nrap(L2,H,S),
S>1,
transform(T,L2,R).
Shall I be pure or impure? Why even consider sacrificing logical-purity if we can save it easily!
Using memberd_t/3 and if_/3, we define list_rset/2 and its left "twin" list_lset/2:
list_rset([], []). % keep rightmost occurrences
list_rset([E|Es], Rs0) :-
if_(memberd_t(E, Es),
Rs0 = Rs,
Rs0 = [E|Rs]),
list_rset(Es, Rs).
list_lset([], []). % keep leftmost occurrences
list_lset([E|Es], Ls) :-
post_pre_lset(Es, [E], Ls). % uses internal auxilary predicate
post_pre_lset([], _, []).
post_pre_lset([E|Es], Pre, Ls0) :- % 2nd arg: look-behind accumulator
if_(memberd_t(E, Pre),
Ls0 = Ls,
Ls0 = [E|Ls]),
post_pre_lset(Es, [E|Pre], Ls).
Let's run some queries!
?- _Es = [1,2,3,1,2], list_lset(_Es, Ls), list_rset(_Es, Rs).
Ls = [1,2,3], Rs = [3,1,2]. % succeeds deterministically
In above query 1 precedes 2 both at the beginning and at the end of the list [1,2,3,1,2]. What if 1 precedes 2 at the beginning but follows it at the end (e.g., [1,2,3,2,1])?
?- _Es = [1,2,3,2,1], list_lset(_Es, Ls), list_rset(_Es, Rs).
Ls = [1,2,3], Rs = [3,2,1]. % succeeds deterministically
Next, we look at a more general list_rset/2 goal that uses a list containing variables only. Thanks to #PauloMoura for his suggestion!
?- Es = [A,B,C,A,B], list_rset(Es,Rs).
Es = [C,C,C,C,C], Rs = [ C], A=B , B=C
; Es = [B,B,C,B,B], Rs = [C, B], A=B , dif(B,C)
; Es = [C,B,C,C,B], Rs = [ C,B], A=C , dif(B,C)
; Es = [A,C,C,A,C], Rs = [ A,C], dif(A,C), B=C
; Es = [A,B,C,A,B], Rs = [C,A,B], dif(A,B), dif(A,C), dif(B,C).
What's up with the residual goals (above)?
Without sufficient instantiation, dif/2 is not decidable.
To save logical soundness, the execution of the prolog-dif constraints is delayed.
Last, one more use-case: an "input" list Xs that has both variables and ground terms.
?- Es = [A,B,z], list_rset(Es,Rs).
Es = [z,z,z], Rs = [ z], A=B , B=z
; Es = [B,B,z], Rs = [B, z], A=B , dif(B,z)
; Es = [z,B,z], Rs = [ B,z], A=z , dif(B,z)
; Es = [A,z,z], Rs = [A, z], dif(A,z), B=z
; Es = [A,B,z], Rs = [A,B,z], dif(A,B), dif(A,z), dif(B,z).
This is a follow-up to this previous answer... In this answer we use dcg!
We build lset//1 upon memberd_t/3 and if_//3—the dcg analogue of if_/3:
lset([]) -->
[].
lset([X|Xs]) -->
[X],
lset_pre(Xs,[X]).
lset_pre([],_) -->
[].
lset_pre([X|Xs],Pre) -->
if_(memberd_t(X,Pre), [], [X]),
lset_pre(Xs,[X|Pre]).
Same for rset//1:
rset([]) -->
[].
rset([X|Xs]) -->
if_(memberd_t(X,Xs), [], [X]),
rset(Xs).
Some sample queries:
?- _Es = [1,2,3,1,2], phrase(lset(_Es),Ls), phrase(rset(_Es),Rs).
Ls = [1,2,3], Rs = [3,1,2]. % succeeds deterministically
?- _Es = [1,2,3,2,1], phrase(lset(_Es),Ls), phrase(rset(_Es),Rs).
Ls = [1,2,3], Rs = [3,2,1]. % succeeds deterministically
This is easier than you are making it. Since the elements in the "set" have to be in the order of last appearance, you don't need to keep a copy of the list at all: just compare to the remainder of the list (the tail).
If you know that the first list is always going to be ground (all elements are integers, for example), you could write:
list_set([], []).
list_set([X|Xs], Ys0) :-
( memberchk(X, Xs)
-> Ys0 = Ys
; Ys0 = [X|Ys]
),
list_set(Xs, Ys).
memberchk/2 can be used to check if a ground term is in a list of ground terms. It will succeed or fail exactly once.
A more general solution is to pose a constraint that an element should be in the set if it is different from all the elements following it, and be dropped otherwise:
list_set([], []).
list_set([X|Xs], [X|Ys]) :-
maplist(dif(X), Xs),
list_set(Xs, Ys).
list_set([X|Xs], Ys) :-
\+ maplist(dif(X), Xs),
list_set(Xs, Ys).
Here, maplist(dif(X), Xs) means:
X is different from every element in the list Xs (the tail).
and \+ Goal succeeds then Goal does not succeed.
With this defintion:
?- list_set([1,2,3,1,2], S).
S = [3, 1, 2] ;
false.
?- list_set([1,2,3,3,1,1,2], S).
S = [3, 1, 2] ;
false.
?- list_set([A,B,C,A,B],Xs).
Xs = [C, A, B],
dif(A, B),
dif(C, B),
dif(C, A) ;
false.

Numbers in a list smaller than a given number

xMenores(_,[],[]).
xMenores(X,[H|T],[R|Z]) :-
xMenores(X,T,Z),
X > H,
R is H.
xMenores takes three parameters:
The first one is a number.
The second is a list of numbers.
The third is a list and is the variable that will contain the result.
The objective of the rule xMenores is obtain a list with the numbers of the list (Second parameter) that are smaller than the value on the first parameter. For example:
?- xMenores(3,[1,2,3],X).
X = [1,2]. % expected result
The problem is that xMenores returns false when X > H is false and my programming skills are almost null at prolog. So:
?- xMenores(4,[1,2,3],X).
X = [1,2,3]. % Perfect.
?- xMenores(2,[1,2,3],X).
false. % Wrong! "X = [1]" would be perfect.
I consider X > H, R is H. because I need that whenever X is bigger than H, R takes the value of H. But I don't know a control structure like an if or something in Prolog to handle this.
Please, any solution? Thanks.
Using ( if -> then ; else )
The control structure you might be looking for is ( if -> then ; else ).
Warning: you should probably swap the order of the first two arguments:
lessthan_if([], _, []).
lessthan_if([X|Xs], Y, Zs) :-
( X < Y
-> Zs = [X|Zs1]
; Zs = Zs1
),
lessthan_if(Xs, Y, Zs1).
However, if you are writing real code, you should almost certainly go with one of the predicates in library(apply), for example include/3, as suggested by #CapelliC:
?- include(>(3), [1,2,3], R).
R = [1, 2].
?- include(>(4), [1,2,3], R).
R = [1, 2, 3].
?- include(<(2), [1,2,3], R).
R = [3].
See the implementation of include/3 if you want to know how this kind of problems are solved. You will notice that lessthan/3 above is nothing but a specialization of the more general include/3 in library(apply): include/3 will reorder the arguments and use the ( if -> then ; else ).
"Declarative" solution
Alternatively, a less "procedural" and more "declarative" predicate:
lessthan_decl([], _, []).
lessthan_decl([X|Xs], Y, [X|Zs]) :- X < Y,
lessthan_decl(Xs, Y, Zs).
lessthan_decl([X|Xs], Y, Zs) :- X >= Y,
lessthan_decl(Xs, Y, Zs).
(lessthan_if/3 and lessthan_decl/3 are nearly identical to the solutions by Nicholas Carey, except for the order of arguments.)
On the downside, lessthan_decl/3 leaves behind choice points. However, it is a good starting point for a general, readable solution. We need two code transformations:
Replace the arithmetic comparisons < and >= with CLP(FD) constraints: #< and #>=;
Use a DCG rule to get rid of arguments in the definition.
You will arrive at the solution by lurker.
A different approach
The most general comparison predicate in Prolog is compare/3. A common pattern using it is to explicitly enumerate the three possible values for Order:
lessthan_compare([], _, []).
lessthan_compare([H|T], X, R) :-
compare(Order, H, X),
lessthan_compare_1(Order, H, T, X, R).
lessthan_compare_1(<, H, T, X, [H|R]) :-
lessthan_compare(T, X, R).
lessthan_compare_1(=, _, T, X, R) :-
lessthan_compare(T, X, R).
lessthan_compare_1(>, _, T, X, R) :-
lessthan_compare(T, X, R).
(Compared to any of the other solutions, this one would work with any terms, not just integers or arithmetic expressions.)
Replacing compare/3 with zcompare/3:
:- use_module(library(clpfd)).
lessthan_clpfd([], _, []).
lessthan_clpfd([H|T], X, R) :-
zcompare(ZOrder, H, X),
lessthan_clpfd_1(ZOrder, H, T, X, R).
lessthan_clpfd_1(<, H, T, X, [H|R]) :-
lessthan_clpfd(T, X, R).
lessthan_clpfd_1(=, _, T, X, R) :-
lessthan_clpfd(T, X, R).
lessthan_clpfd_1(>, _, T, X, R) :-
lessthan_clpfd(T, X, R).
This is definitely more code than any of the other solutions, but it does not leave behind unnecessary choice points:
?- lessthan_clpfd(3, [1,3,2], Xs).
Xs = [1, 2]. % no dangling choice points!
In the other cases, it behaves just as the DCG solution by lurker:
?- lessthan_clpfd(X, [1,3,2], Xs).
Xs = [1, 3, 2],
X in 4..sup ;
X = 3,
Xs = [1, 2] ;
X = 2,
Xs = [1] ;
X = 1,
Xs = [] .
?- lessthan_clpfd(X, [1,3,2], Xs), X = 3. %
X = 3,
Xs = [1, 2] ; % no error!
false.
?- lessthan_clpfd([1,3,2], X, R), R = [1, 2].
X = 3,
R = [1, 2] ;
false.
Unless you need such a general approach, include(>(X), List, Result) is good enough.
This can also be done using a DCG:
less_than([], _) --> [].
less_than([H|T], N) --> [H], { H #< N }, less_than(T, N).
less_than(L, N) --> [H], { H #>= N }, less_than(L, N).
| ?- phrase(less_than(R, 4), [1,2,3,4,5,6]).
R = [1,2,3] ? ;
You can write your predicate as:
xMenores(N, NumberList, Result) :- phrase(less_than(Result, N), NumberList).
You could write it as a one-liner using findall\3:
filter( N , Xs , Zs ) :- findall( X, ( member(X,Xs), X < N ) , Zs ) .
However, I suspect that the point of the exercise is to learn about recursion, so something like this would work:
filter( _ , [] , [] ) .
filter( N , [X|Xs] , [X|Zs] ) :- X < N , filter(N,Xs,Zs) .
filter( N , [X|Xs] , Zs ) :- X >= N , filter(N,Xs,Zs) .
It does, however, unpack the list twice on backtracking. An optimization here would be to combine the 2nd and 3rd clauses by introducing a soft cut like so:
filter( _ , [] , [] ) .
filter( N , [X|Xs] , [X|Zs] ) :-
( X < N -> Zs = [X|Z1] ; Zs = Z1 ) ,
filter(N,Xs,Zs)
.
(This is more like a comment than an answer, but too long for a comment.)
Some previous answers and comments have suggested using "if-then-else" (->)/2 or using library(apply) meta-predicate include/3. Both methods work alright, as long as only plain-old Prolog arithmetics—is/2, (>)/2, and the like—are used ...
?- X = 3, include(>(X),[1,3,2,5,4],Xs).
X = 3, Xs = [1,2].
?- include(>(X),[1,3,2,5,4],Xs), X = 3.
ERROR: >/2: Arguments are not sufficiently instantiated
% This is OK. When instantiation is insufficient, an exception is raised.
..., but when doing the seemingly benign switch from (>)/2 to (#>)/2, we lose soundness!
?- X = 3, include(#>(X),[1,3,2,5,4],Xs).
X = 3, Xs = [1,2].
?- include(#>(X),[1,3,2,5,4],Xs), X = 3.
false.
% This is BAD! Expected success with answer substitutions `X = 3, Xs = [1,2]`.
No new code is presented in this answer.
In the following we take a detailed look at different revisions of this answer by #lurker.
Revision #1, renamed to less_than_ver1//2. By using dcg and clpfd, the code is both very readable and versatile:
less_than_ver1(_, []) --> [].
less_than_ver1(N, [H|T]) --> [H], { H #< N }, less_than_ver1(N, T).
less_than_ver1(N, L) --> [H], { H #>= N }, less_than_ver1(N, L).
Let's query!
?- phrase(less_than_ver1(N,Zs),[1,2,3,4,5]).
N in 6..sup, Zs = [1,2,3,4,5]
; N = 5 , Zs = [1,2,3,4]
; N = 4 , Zs = [1,2,3]
; N = 3 , Zs = [1,2]
; N = 2 , Zs = [1]
; N in inf..1, Zs = []
; false.
?- N = 3, phrase(less_than_ver1(N,Zs),[1,2,3,4,5]).
N = 3, Zs = [1,2] % succeeds, but leaves useless choicepoint
; false.
?- phrase(less_than_ver1(N,Zs),[1,2,3,4,5]), N = 3.
N = 3, Zs = [1,2]
; false.
As a small imperfection, less_than_ver1//2 leaves some useless choicepoints.
Let's see how things went with the newer revision...
Revision #3, renamed to less_than_ver3//2:
less_than_ver3([],_) --> [].
less_than_ver3(L,N) --> [X], { X #< N -> L=[X|T] ; L=T }, less_than_ver3(L,N).
This code uses the if-then-else ((->)/2 + (;)/2) in order to improve determinism.
Let's simply re-run the above queries!
?- phrase(less_than_ver3(Zs,N),[1,2,3,4,5]).
N in 6..sup, Zs = [1,2,3,4,5]
; false. % all other solutions are missing!
?- N = 3, phrase(less_than_ver3(Zs,N),[1,2,3,4,5]).
N = 3, Zs = [1,2] % works as before, but no better.
; false. % we still got the useless choicepoint
?- phrase(less_than_ver3(Zs,N),[1,2,3,4,5]), N = 3.
false. % no solution!
% we got one with revision #1!
Surprise! Two cases that worked before are now (somewhat) broken, and the determinism in the ground case is no better... Why?
The vanilla if-then-else often cuts too much too soon, which is particularly problematic with code which uses coroutining and/or constraints.
Note that (*->)/2 (a.k.a. "soft-cut" or if/3), fares only a bit better, not a lot!
As if_/3 never ever cuts more (often than) the vanilla if-then-else (->)/2, it cannot be used in above code to improve determinism.
If you want to use if_/3 in combination with constraints, take a step back and write code that is non-dcg as the first shot.
If you're lazy like me, consider using a meta-predicate like tfilter/3 and (#>)/3.
This answer by #Boris presented a logically pure solution which utilizes clpfd:zcompare/3 to help improve determinism in certain (ground) cases.
In this answer we will explore different ways of coding logically pure Prolog while trying to avoid the creation of useless choicepoints.
Let's get started with zcompare/3 and (#<)/3!
zcompare/3 implements three-way comparison of finite domain variables and reifies the trichotomy into one of <, =, or >.
As the inclusion criterion used by the OP was a arithmetic less-than test, we propose using
(#<)/3 for reifying the dichotomy into one of true or false.
Consider the answers of the following queries:
?- zcompare(Ord,1,5), #<(1,5,B).
Ord = (<), B = true.
?- zcompare(Ord,5,5), #<(5,5,B).
Ord = (=), B = false.
?- zcompare(Ord,9,5), #<(9,5,B).
Ord = (>), B = false.
Note that for all items to be selected both Ord = (<) and B = true holds.
Here's a side-by-side comparison of three non-dcg solutions based on clpfd:
The left one uses zcompare/3 and first-argument indexing on the three cases <, =, and >.
The middle one uses (#<)/3 and first-argument indexing on the two cases true and false.
The right one uses (#<)/3 in combination with if_/3.
Note that we do not need to define auxiliary predicates in the right column!
less_than([],[],_). % less_than([],[],_). % less_than([],[],_).
less_than([Z|Zs],Ls,X) :- % less_than([Z|Zs],Ls,X) :- % less_than([Z|Zs],Ls,X) :-
zcompare(Ord,Z,X), % #<(Z,X,B), % if_(Z #< X,
ord_lt_(Ord,Z,Ls,Rs), % incl_lt_(B,Z,Ls,Rs), % Ls = [Z|Rs],
less_than(Zs,Rs,X). % less_than(Zs,Rs,X). % Ls = Rs),
% % less_than(Zs,Rs,X).
ord_lt_(<,Z,[Z|Ls],Ls). % incl_lt_(true ,Z,[Z|Ls],Ls). %
ord_lt_(=,_, Ls ,Ls). % incl_lt_(false,_, Ls ,Ls). %
ord_lt_(>,_, Ls ,Ls). % %
Next, let's use dcg!
In the right column we use if_//3 instead of if_/3.
Note the different argument orders of dcg and non-dcg solutions: less_than([1,2,3],Zs,3) vs phrase(less_than([1,2,3],3),Zs).
The following dcg implementations correspond to above non-dcg codes:
less_than([],_) --> []. % less_than([],_) --> []. % less_than([],_) --> [].
less_than([Z|Zs],X) --> % less_than([Z|Zs],X) --> % less_than([Z|Zs],X) -->
{ zcompare(Ord,Z,X) }, % { #<(Z,X,B) }, % if_(Z #< X,[Z],[]),
ord_lt_(Ord,Z), % incl_lt_(B,Z), % less_than(Zs,X).
less_than(Zs,X). % less_than(Zs,X). %
% %
ord_lt_(<,Z) --> [Z]. % incl_lt_(true ,Z) --> [Z]. %
ord_lt_(=,_) --> []. % incl_lt_(false,_) --> []. %
ord_lt_(>,_) --> []. % %
OK! Saving the best for last... Simply use meta-predicate tfilter/3 together with (#>)/3!
less_than(Xs,Zs,P) :-
tfilter(#>(P),Xs,Zs).
The dcg variant in this previous answer is our starting point.
Consider the auxiliary non-terminal ord_lt_//2:
ord_lt_(<,Z) --> [Z].
ord_lt_(=,_) --> [].
ord_lt_(>,_) --> [].
These three clauses can be covered using two conditions:
Ord = (<): the item should be included.
dif(Ord, (<)): it should not be included.
We can express this "either-or choice" using if_//3:
less_than([],_) --> [].
less_than([Z|Zs],X) -->
{ zcompare(Ord,Z,X) },
if_(Ord = (<), [Z], []),
less_than(Zs,X).
Thus ord_lt_//2 becomes redundant.
Net gain? 3 lines-of-code !-)

Prolog delete: doesn't delete all elements that unify with Element

I'm having an issue with SWI-Prolog's delete/3 predicate.
The easiest way is just a quick example:
?- delete([(1,1),(1,2),(3,2)], (1,_), List).
List = [(1,2),(3,2)].
I would expect (1,2) to also be deleted, since (1,_) unifies with (1,2). The SWIPL help says:
Delete all members of List1 that simultaneously unify with Elem and unify the result with List2.
Why is this and how can I delete everything that unifies with (1,_)?
" Delete all members of List1 that simultaneously unify with Elem and unify the result with List2."
(1,X) first unifies with (1,1). therefore, X is unified with 1 and cannot be unified with 2 to delete (1,2).
so the problem is not that it does not delete all of the members; it's that it doesnt unify simultaneously with (1,2) and (1,1)
(try delete([(1,1),(1,2),(1,1),(3,2)],(1,_),List).
btw, according to the swi-prolog manual:
delete(?List1, ?Elem, ?List2)
Is true when Lis1, with all occurences of Elem deleted results in List2.
also, delete/3 is deprecated:
There are too many ways in which one might want to delete elements from a list to justify the name.
Think of matching (= vs. ==), delete first/all, be deterministic or not.
So the easiest way is to write your own predicate. Something like:
my_delete(Pattern,[Pattern|T],TD):-
my_delete(Pattern,T,TD).
my_delete(Pattern,[H|T],[H|TD]):-
my_delete(Pattern,T,TD).
perhaps?
check exclude/3, include/3, partition/4
Use meta-predicate texclude/3 in combination with the
reified term equality predicate
(=)/3!
First, we try using (=)/3 directly...
?- texclude(=((1,V)), [(1,1),(1,2),(3,2)], KVs).
KVs = [ (1,2),(3,2)], V=1 ;
KVs = [(1,1), (3,2)], V=2 ;
KVs = [(1,1),(1,2),(3,2)], dif(V,1), dif(V,2).
Not quite! For our next tries we are going to use lambda expressions.
:- use_module(library(lambda)).
Let's query---once with texclude/3, once with tinclude/3, and once with tpartition/4:
?- texclude( \ (K,_)^(K=1), [(1,1),(1,2),(3,2)], Fs).
Fs = [(3,2)]. % succeeds deterministically
?- tinclude( \ (K,_)^(K=1), [(1,1),(1,2),(3,2)], Ts).
Ts = [(1,1),(1,2)]. % succeeds deterministically
?- tpartition(\ (K,_)^(K=1), [(1,1),(1,2),(3,2)], Ts,Fs).
Ts = [(1,1),(1,2)], Fs = [(3,2)]. % succeeds deterministically
Alright! Do we get the same solutions if the list items are bound after the texclude/3 call?
?- texclude(\ (K,_)^(K=1), [A,B,C], Fs), A = (1,1), B = (1,2), C = (3,2).
A = (1,1), B = (1,2), C = (3,2), Fs = [(3,2)] ; % succeeds with choice point
false.
Yes! At last, consider the following quite general query:
?- texclude(\ (K,_)^(K=1), [A,B], Fs).
Fs = [ ], A = ( 1,_A1), B = ( 1,_B1) ;
Fs = [ B], A = ( 1,_A1), B = (_B0,_B1), dif(_B0,1) ;
Fs = [A ], A = (_A0,_A1), B = ( 1,_B1), dif(_A0,1) ;
Fs = [A,B], A = (_A0,_A1), B = (_B0,_B1), dif(_A0,1), dif(_B0,1).
Note that above goals restrict all list items to have the form (_,_). Thus the following query fails:
?- texclude(\ (K,_)^(K=1), [x,_], _).
false.
This answer tries to generalize the idea presented in previous answer.
Let's define a reified variant of subsumes_term/2:
list_nonvardisj([A],C) :-
!,
C = nonvar(A).
list_nonvardisj([A|As],(nonvar(A);C)) :-
list_nonvardisj(As,C).
subsumes_term_t(General,Specific,Truth) :-
subsumes_term(General,Specific),
!,
term_variables(General,G_vars),
free4evrs(G_vars),
Truth = true.
subsumes_term_t(General,Specific,Truth) :-
Specific \= General,
!,
Truth = false.
subsumes_term_t(General,Specific,Truth) :-
term_variables(Specific,S_vars),
( S_vars = [V]
-> freeze(V,subsumes_term_t(General,Specific,Truth))
; S_vars = [_|_]
-> list_nonvardisj(S_vars,S_wakeup),
when(S_wakeup,subsumes_term_t(General,Specific,Truth))
; throw(error(instantiation_error, subsumes_term_t/3))
),
( Truth = true
; Truth = false
).
The above definition of the reified predicate subsumes_term_t/3 uses free4evrs/1 to ensure that the "generic" term passed to subsumes_term/2 is not instantiated any further.
For SICStus Prolog, we can define it as follows:
:- module(free4evr,[free4evr/1,free4evrs/1]).
:- use_module(library(atts)).
:- attribute nvrb/0. % nvrb ... NeVeR Bound
verify_attributes(V,_,Goals) :-
get_atts(V,nvrb),
!,
Goals = [throw(error(uninstantiation_error(V),free4evr/1))].
verify_attributes(_,_,[]).
attribute_goal(V,free4evr(V)) :-
get_atts(V,nvrb).
free4evr(V) :-
nonvar(V),
!,
throw(error(uninstantiation_error(V),free4evr/1)).
free4evr(V) :-
( get_atts(V,nvrb)
-> true
; put_atts(Fresh,nvrb),
V = Fresh
).
free4evrs([]).
free4evrs([V|Vs]) :-
free4evr(V),
free4evrs(Vs).
Let's put subsumes_term_t/3 to use!
?- texclude(subsumes_term_t(1-X), [A,B,C], Fs), A = 1-1, B = 1-2, C = 3-2.
A = 1-1, B = 1-2, C = 3-2, Fs = [C], free4evr(X) ? ; % succeeds with choice-point
no
?- texclude(subsumes_term_t(1-X), [x,1-Y,2-3], Fs).
Fs = [x,2-3], free4evr(X) ? ;
no
What happens if we instantiate variable X in above query sometime after the call to texclude/3?
?- texclude(subsumes_term_t(1-X), [x,1-Y,2-3], Fs), X=something.
! error(uninstantiation_error(something),free4evr/1)