Related
How to get the LCM of a list in Prolog?
Lets say the list is: [1,2,3,4,5] and the LCM will be 60.
I have the following code fo the GCD and the LCM that work for 2 numbers but dont know how to apply to lists.
gcd(X, 0, X) :- !.
gcd(X, Y, Z) :-
H is X rem Y,
gcd(Y, H, Z).
lcm(X,Y,LCM):-
gcd(X,Y,GCD),
LCM is X*Y//GCD.
gcd_of_list([], 0) :- !.
gcd_of_list([X|Xs], GCD) :- gcd_of_list(Xs, GCD2), gcd(X, GCD2, GCD).
?- gcd_of_list([150,1000,120], GCD).
GCD = 10.
lcm_of_list([],1) :- !.
lcm_of_list([X|Xs],LCM) :- lcm_of_list(Xs,LCM2), lcm(X,LCM2,LCM).
?- lcm_of_list([9, 7, 10, 9, 7, 8, 5, 10, 1],LCM).
LCM = 2520.
BTW, an interesting result is that you cannot use gcd_of_list to compute lcm_of_list directly, for example:
%% A wrong attempt
lcm_of_list(Lst, LCM) :-
gcd_of_list(Lst, GCD),
mul_list(Lst, Mul),
LCM is Mul // GCD.
See https://math.stackexchange.com/a/319310/430364
There can be no formula that computes lcm(a,b,c) using only the values of abc and gcd(a,b,c) as input.
Therefore, we must compute lcm_of_list from scratch.
Given a list of possible summands I want to determine which, if any, can form a given sum. For example, with [1,2,3,4,5] I can make the sum of 9 with [4,5], [5,3,1], and [4,3,2].
I am using GNU Prolog and have something like the following which does not work
numbers([1,2,3,4,5]).
all_unique(_, []).
all_unique(L, [V|T]) :-
fd_exactly(1, L, V),
all_unique(L, T).
fd_sum([], Sum).
fd_sum([H|T], Sum):-
S = Sum + H,
fd_sum(T, S).
sum_clp(N, Summands):-
numbers(Numbers),
length(Numbers, F),
between(1, F, X),
length(S, X),
fd_domain(S, Numbers),
fd_domain(Y, [N]),
all_unique(S, Numbers),
fd_sum(S, Sum),
Sum #= Y,
fd_labeling(S).
I think the main problem is that I am not representing the constraint on the sum properly? Or maybe it is something else?
Just in case you're really interested in CLP(FD), here is your corrected program.
numbers([1,2,3,4,5]).
% note: use builtins where available, both for efficiency and correctness
%all_unique(_, []).
%all_unique(L, [V|T]) :-
% fd_exactly(1, L, V),
% all_unique(L, T).
fd_sum([], 0). % sum_fd_SO.pl:8: warning: singleton variables [Sum] for fd_sum/2
fd_sum([H|T], Sum):-
% note: use CLP(FD) operators and the correct operands
Sum #= S + H,
fd_sum(T, S).
sum_clp(N, S):- % sum_fd_SO.pl:13-23: warning: singleton variables [Summands] for sum_clp/2
numbers(Numbers),
length(Numbers, F),
between(1, F, X),
length(S, X),
fd_domain(S, Numbers),
%fd_domain(Y, [N]),
%all_unique(S, Numbers),
fd_all_different(S),
fd_sum(S, N),
%Sum #= Y,
fd_labeling(S).
test
?- sum_clp(3,L).
L = [3] ? ;
L = [1,2] ? ;
L = [2,1] ? ;
no
I think mixing the code for sublist into clp code is causing some confusion. GNU-Prolog has a sublist/2 predicate, you can use that.
You seem to be building the arithmetic expression with fd_sum but it is incorrectly implemented.
sum_exp([], 0).
sum_exp([X|Xs], X+Xse) :-
sum_exp(Xs, Xse).
sum_c(X, N, Xsub) :-
sublist(Xsub, X),
sum_exp(Xsub, Xe),
N #= Xe.
| ?- sum_exp([A, B, C, D], X).
X = A+(B+(C+(D+0)))
yes
| ?- sum_c([1, 2, 3, 4, 5], 9, X).
X = [4,5] ? ;
X = [2,3,4] ? ;
X = [1,3,5] ? ;
(1 ms) no
| ?- length(X, 4), sum_c(X, 4, [A, B]), member(A, [1, 2, 3]).
A = 1
B = 3
X = [_,_,1,3] ? ;
A = 2
B = 2
X = [_,_,2,2] ? ;
A = 3
B = 1
X = [_,_,3,1] ?
yes
The following Prolog program defines a predicate sorted/2 for sorting by permutation (permutation sort) in ascending order a list passed in first argument, which results in the list passed in second argument:
sorted(X, Y) :-
permuted(X, Y),
ordered(Y).
permuted([], []).
permuted(U, [V|W]) :-
permuted(X, W),
deleted(V, U, X).
deleted(X, [X|Y], Y).
deleted(U, [V|W], [V|X]) :-
deleted(U, W, X).
ordered([]).
ordered([_]).
ordered([X, Y|Z]) :-
ordered([Y|Z]), X =< Y.
How to solve the following issues?
The program duplicates solutions for queries in which a list with duplicate elements is passed in second argument:
?- sorted(X, [1, 1, 2]).
X = [1, 1, 2]
; X = [1, 1, 2]
; X = [1, 2, 1]
; X = [1, 2, 1]
; X = [2, 1, 1]
; X = [2, 1, 1]
; false.
The program exhausts resources for queries in which a free variable is passed in second argument:
?- sorted([2, 1, 1], Y).
Y = [1, 1, 2]
; Y = [1, 1, 2]
;
Time limit exceeded
The Prolog program is based on the Horn clause program given at section 11 of Robert Kowalski’s famous paper Predicate Logic as Programming Language:
To solve non-termination, you can add same_length/2 to sorted/2 as #false suggested:
sorted(X, Y) :-
same_length(X, Y),
permuted(X, Y),
ordered(Y).
same_length([], []).
same_length([_|Xs], [_|Ys]) :-
same_length(Xs, Ys).
Or you may embed it into permuted/2 by adding a new argument:
sorted(X, Y) :-
permuted(X, X, Y),
ordered(Y).
permuted([], [], []).
permuted(U, [_|L1], [V|W]) :-
permuted(X, L1, W),
deleted(V, U, X).
The program will still return duplicates as it only sees one item at a time.
To solve duplication, you can either generate all permutations and discard the repeated ones (which is not efficient), or only generate distinct permutations. The following modification does the latter by taking the idea of the recursive procedure permuted/2 + deleted/2 which for each item puts it at the beginning of the list and does a recursive call on the remaining list, and changes it to another recursive procedure permuted_all/2 + deleted_all/2 which for each group of same items puts them at the beginning of the list and does a recursive call on the remaining list. This program uses difference lists for better efficiency:
sorted(X, Y) :-
same_length(X, Y),
permuted_all(X, Y),
ordered(Y).
permuted_all([], []).
permuted_all(U, [V|W]) :-
deleted_all(V, U, X, n-T, [V|W]),
permuted_all(X, T).
% deleted_all(Item, List, Remainder, n-T, Items|T)
deleted_all(_, [], [], y-[X|Xs], [X|Xs]).
deleted_all(X, [V|Y], [V|Y1], y-[X|Xs], Xs1) :-
dif(X, V),
deleted_all(X, Y, Y1, y-[X|Xs], Xs1).
deleted_all(X, [X|Y], Y1, _-Xs, Xs1) :-
deleted_all(X, Y, Y1, y-[X|Xs], Xs1).
deleted_all(U, [V|W], [V|X], n-T, Xs) :-
dif(U, V),
deleted_all(U, W, X, n-T, Xs).
Sample runs:
?- sorted(X, [1, 1, 2]).
X = [1, 2, 1]
; X = [1, 1, 2]
; X = [2, 1, 1]
; false.
?- sorted([2, 1, 1], Y).
Y = [1, 1, 2]
; false.
As per OPs comment asking for a version which does not use difference lists, here goes one which instead obtains the remainder using same_length/2 + append/3 and with added comments:
permuted_all([], []).
permuted_all(U, [V|W]) :-
deleted_all(V, U, X, n, [V|W]),
same_length(X, T), % the remaining list X has the same length as T
append(_, T, [V|W]), % T corresponds to the last items of [V|W]
permuted_all(X, T). % T is a permutation of X
% deleted_all(Item, List, Remainder, n, Items|_)
deleted_all(_, [], [], y, _). % base case
deleted_all(X, [V|Y], [V|Y1], y, Xs1) :-
% recursive step when the current item is not the one we are gathering
dif(X, V),
deleted_all(X, Y, Y1, y, Xs1).
deleted_all(X, [X|Y], Y1, _, [X|Xs1]) :-
% recursive step when the current item is the one we are gathering
deleted_all(X, Y, Y1, y, Xs1).
deleted_all(U, [V|W], [V|X], n, Xs) :-
% recursive step when we have not selected yet the item we will be gathering
dif(U, V),
deleted_all(U, W, X, n, Xs).
Your second problem can by solved by replacing first line with
sorted(X, Y) :-
permuted(X, Y),
ordered(Y),
!.
or
sorted(X, Y) :-
permuted(X, Y),
ordered(Y),
length(X, Z),
length(Y, Z).
The first one is not so easy to solve because of the implementation of this algorithm. Both 1st [1, 1, 2] and 2nd [1, 1, 2] are valid permutations since your code that generated permutations generates all permutations not unique permutations.
Say you have the following predicate:
random_int(X/Y):-
random(1,100,X),
random(1,100,Y),
X\=Y.
How can I populate a list of size n using the result of this predicate?
I tried the following code but it only populates the list if random_int(X) is true at the first attempt, i.e. it does not backtrack to try other combinations of X and Y.
findall(X,(between(1,N,_), random_int(X)),L).
I find this small 'application' of clpfd interesting:
?- N=10,M=12, repeat, findall(X, (between(1,N,_),random(1,M,X)), L), clpfd:all_different(L).
N = 10,
M = 12,
L = [5, 4, 6, 7, 9, 11, 2, 3, 8|...]
.
note: M must be > N
I guess a simple way to do it is to make a list of 1:100, and draw 100 times from it a sample of size 2, without replacement. Since this is Prolog and not R, you can instead do:
:- use_module(library(lists)).
:- use_module(library(random)).
random_pairs(Pairs) :-
findall(X/Y,
( between(1, 100, _),
randseq(2, 100, [X,Y])
), R).
This is available in SWI-Prolog at least, but it is free software and the source to randseq/3 is available on the web site.
And since it's better to not use findall unless strictly necessary, it would probable better to write:
random_pairs(Pairs) :-
length(Pairs, 100),
maplist(randseq(2, 100), Pairs).
or, if the X/Y is important,
random_pairs(Pairs) :-
length(Pairs, 100),
maplist(rand_couple(100), Pairs).
rand_couple(N, X/Y) :-
randseq(2, N, [X,Y]).
TL;DR Use the available libraries
You could do it with findall/3:
random_list(N, L) :-
findall(X, (between(1,N,_), random(50,100,X)), L).
Another tidy way to do this would be:
random_list(N, L) :-
length(L, N),
maplist(random(50, 100), L).
Which results in:
| ?- random_list(5, L).
L = [69,89,89,95,59]
yes
| ?-
In general, if you have a predicate, p(X1,X2,...,Xn,Y), and a list you want to fill with result Y using successive calls to p/(n+1), you can use length(List, Length) to set the length of your list, and then maplist(p(X1,...,Xn), List) to populate the list. Or, using the findall/3, you can do findall(X, (between(1,N,_), p(X1,...,Xn,X)), L)..
EDIT based upon the updated conditions of the question that the generated list be unique values...
The random predicates are not generators, so they don't create new random numbers on backtracking (either unique or otherwise). So this solution, likewise, will generate one list which meets the requirements, and then just succeed without generating more such lists on backtracking:
% Generate a random number X between A and B which is not in L
rand_not_in(A, B, L, X) :-
random(A, B, X1),
( memberchk(X1, L)
-> rand_not_in(A, B, L, X)
; X = X1
).
% Generate a list L of length N consisting of unique random numbers
% between A and B
random_list(N, L) :-
random_list(N, 50, 100, [], L).
random_list(N, A, B, Acc, L) :-
N > 0,
rand_not_in(A, B, A, X),
N1 is N - 1,
random_list(N1, A, B, [X|A], L).
random_list(0, _, _, L, L).
Yet another approach, in SWI Prolog, you can use randseq, which will give a random sequence in a range 1 to N. Just scale it:
random_list(N, A, B, L) :-
A < B,
Count is B - A + 1,
randseq(N, Count, L1),
Offset is A - 1,
maplist(offset(Offset), L1, L).
offset(X, Offset, Y) :-
Y is X + Offset.
?- random_list(5, 50, 100, L).
L = [54, 91, 90, 78, 75].
?-
random_len([],0).
random_len([Q|T],N) :-
random(1,100,Q),
random_len(T,X),
N is X+1.
How do you get the product of a list from left to right?
For example:
?- product([1,2,3,4], P).
P = [1, 2, 6, 24] .
I think one way is to overload the functor and use 3 arguments:
product([H|T], Lst) :- product(T, H, Lst).
I'm not sure where to go from here.
You can use library(lambda) found here : http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/lambda.pl
Quite unreadable :
:- use_module(library(lambda)).
:- use_module(library(clpfd)).
product(L, R) :-
foldl(\X^Y^Z^(Y = []
-> Z = [X, [X]]
; Y = [M, Lst],
T #= X * M,
append(Lst, [T], Lst1),
Z = [T, Lst1]),
L, [], [_, R]).
Thanks to #Mike_Hartl for his advice, the code is much simple :
product([], []).
product([H | T], R) :-
scanl(\X^Y^Z^( Z #= X * Y), T, H, R).
seems like a list copy, just multiplying by last element handled. Let's start from 1 for the leftmost element:
product(L, P) :-
product(L, 1, P).
product([X|Xs], A, [Y|Ys]) :-
Y is X * A,
product(Xs, Y, Ys).
product([], _, []).
if we use library(clpfd):
:- [library(clpfd)].
product([X|Xs], A, [Y|Ys]) :-
Y #= X * A,
product(Xs, Y, Ys).
product([], _, []).
it works (only for integers) 'backward'
?- product(L, [1,2,6,24]).
L = [1, 2, 3, 4].
Probably very dirty solution (I am new to Prolog):
product([ListHead|ListTail], Answer) :-
product_acc(ListTail, [ListHead], Answer).
product_acc([ListHead|ListTail], [AccHead|AccTail], Answer) :-
Product is ListHead * AccHead,
append([Product, AccHead], AccTail, TempList),
product_acc(ListTail, TempList, Answer).
product_acc([], ReversedList, Answer) :-
reverse(ReversedList, Answer).
So basically at the beginning we call another predicate which has
extra "variable" Acc which is accumulator list.
So we take out head (first number) from original list and put it in
to Accumulator list.
Then we always take head (first number) from original list and
multiply it with head (first number) from accumulator list.
Then we have to append our new number which we got by multiplying
with the head from accumulator and later with the tail
Then we call same predicate again until original list becomes empty
and at the end obviously we need to reverse it.
And it seems to work
?- product([1,2,3,4], L).
L = [1, 2, 6, 24].
?- product([5], L).
L = [5].
?- product([5,4,3], L).
L = [5, 20, 60].
Sorry if my explanation is not very clear. Feel free to comment.