% FINITE and INFINITE DOMAINS % 910527 ECRC thom fruehwirth % 910913 modified % 920409 element/3 added % 920616 more CHIP predicates added % 930726 started porting to CHR release % 931014 mult/3 added for CHIC user meeting % 931201 ported to CHR release % 931208 removed special case of integer domain % 940304 element/3 constraint loop fixed % 961017 Christian Holzbaur SICStus mods % 980714 Thom Fruehwirth, some updates reagrding alread_in* % just quick port from Eclipse CHR library version % does not take advantage of Sicstus CHR library features! % Simplifies domains together with inequalities and some more CHIP predicates: % element/3, atmost/3, alldistinct/1, circuit/1 and mult/3 % It also includes paired (!) domains (see element constraint) :- use_module( library(chr)). :- use_module( library('chr/getval')). :- use_module( library(lists), [member/2,last/2]). :- use_module( library(ordsets), [ list_to_ord_set/2, ord_intersection/3 ]). handler domain. option(already_in_store, on). option(already_in_heads, off). % see pragma already_in_heads option(check_guard_bindings, off). % for domain constraints operator(700,xfx,'::'). operator(600,xfx,'..'). operator(600,xfx,':'). % clash with module operator? % for inequality constraints operator(700,xfx,lt). operator(700,xfx,le). operator(700,xfx,gt). operator(700,xfx,ge). operator(700,xfx,ne). % X::Dom - X must be element of the finite or infinite domain Dom % Domains can be either numbers (including arithemtic expressions) % or arbitrary ground terms (!), the domain is set with setval(domain,Kind), % where Kind is either number or term. Default for Kind is term. :- setval(domain,term). % set default % INEQUALITIES =============================================================== % inequalities over numbers (including arithmetic expressions) or terms constraints lt/2,le/2,ne/2. A gt B :- B lt A. % constraints gt/2,ge/2 A ge B :- B le A. % some basic simplifications A lt A <=> fail. A le A <=> true. A ne A <=> fail. A lt B,B lt A <=> fail. A le B,B le A <=> A=B. A ne B \ B ne A <=> true. % for number domain, allow arithmetic expressions in the arguments A lt B <=> domain(number),ground(A),\+ number(A) | A1 is A, A1 lt B. B lt A <=> domain(number),ground(A),\+ number(A) | A1 is A, B lt A1. A le B <=> domain(number),ground(A),\+ number(A) | A1 is A, A1 le B. B le A <=> domain(number),ground(A),\+ number(A) | A1 is A, B le A1. A ne B <=> domain(number),ground(A),\+ number(A) | A1 is A, A1 ne B. B ne A <=> domain(number),ground(A),\+ number(A) | A1 is A, B ne A1. % use built-ins to solve the predicates if arguments are known A lt B <=> ground(A),ground(B) | (domain(number) -> A < B ; A @< B). A le B <=> ground(A),ground(B) | (domain(number) -> A =< B ; A @=< B). A ne B <=> ground(A),ground(B) | (domain(number) -> A =\= B ; A \== B). % FINITE and INFINITE DOMAINS ================================================ constraints (::)/2. % enforce groundness of domain expression X::Dom <=> nonground(Dom) | raise_exception( instantiation_error(X::Dom,2)). constraints labeling/0. labeling, (X::[Y|L]) # Ph <=> member(X,[Y|L]), labeling pragma passive(Ph). % binary search by splitting domain in halves labeling, (X::Min:Max) # Ph <=> domain(number),Min+0.5 % assume we have integer domain Mid is (Min+Max)//2, Next is Mid+1 ; Mid is (Min+Max)/2, Next=Mid % splitted domains overlap at Mid for floats ), ( X::Min:Mid ; X::Next:Max % ; % Min+1>Max, % for floats only, to get X also bound % X=Min % or X=Max etc. ), labeling pragma passive(Ph). nonground(X) :- ground(X), !, fail. nonground(_). domain(Kind) :- getval(domain,Kind). % CHIP list shorthand for domain variables % list must be known (end in the empty list) [X|L]::Dom <=> makedom([X|L],Dom). makedom([],D) :- true. makedom([X|L],D) :- nonvar(L), X::D, makedom(L,D). % Consecutive integer domain --------------------------------------------- % X::Min..Max - X is an integer between the numbers Min and Max (included) % constraint is mapped to enumeration domain constraint X::Min..Max <=> Min0 is Min, (Min0=:=round(float(Min0)) -> Min1 is integer(Min0) ; Min1 is integer(Min0+1)), Max1 is integer(Max), interval(Min1,Max1,L), X::L. interval(M,N,[M|Ns]):- M list_to_ord_set([A|L],SL), SL\==[A|L] | X::SL. % for number domain, allow arithmetic expressions in domain X::[A|L] <=> domain(number), member(X,[A|L]), \+ number(X) | eval_list([A|L],L1),list_to_ord_set(L1,L2), X::L2. eval_list([],[]). eval_list([X|L1],[Y|L2]):- Y is X, eval_list(L1,L2). % special cases X::[] <=> fail. X::[Y] <=> X=Y. X::[A|L] <=> ground(X) | (member(X,[A|L]) -> true). % intersection of domains for the same variable % without pragma already_in_heads, needs already_in_store X::[A1|L1] \ X::[A2|L2] <=> ord_intersection([A1|L1],[A2|L2],L), L \== [A2|L2] | X::L. % interaction with inequalities X::[A|L] \ X ne Y <=> integer(Y), remove(Y,[A|L],L1) | X::L1. X::[A|L] \ Y ne X <=> integer(Y), remove(Y,[A|L],L1) | X::L1. X::[A|L], Y le X ==> ground(Y), remove_lower(Y,[A|L],L1) | X::L1. X::[A|L], X le Y ==> ground(Y), remove_higher(Y,[A|L],L1) | X::L1. X::[A|L], Y lt X ==> ground(Y), remove_lower(Y,[A|L],L1),remove(Y,L1,L2) | X::L2. X::[A|L], X lt Y ==> ground(Y), remove_higher(Y,[A|L],L1),remove(Y,L1,L2) | X::L2. % interaction with interval domain X::[A|L], X::Min:Max ==> remove_lower(Min,[A|L],L1),remove_higher(Max,L1,L2) | X::L2. % propagation of bounds X le Y, Y::[A|L] ==> var(X) | last([A|L],Max), X le Max. X le Y, X::[Min|_] ==> var(Y) | Min le Y. X lt Y, Y::[A|L] ==> var(X) | last([A|L],Max), X lt Max. X lt Y, X::[Min|_] ==> var(Y) | Min lt Y. % Interval domain --------------------------------------------------------- % X::Min:Max - X must be a ground term between Min and Max (included) % for number domain, allow for arithmetic expressions ind omain % for integer domains, X::Min..Max should be used X::Min:Max <=> domain(number), \+ (number(Min),number(Max)) | Min1 is Min, Max1 is Max, X::Min1:Max1. % special cases X::Min:Min <=> X=Min. X::Min:Max <=> (domain(number) -> Min>Max ; Min@>Max) | fail. X::Min:Max <=> ground(X) | (domain(number) -> Min= maximum(Min1,Min2,Min), minimum(Max1,Max2,Max), (Min \== Min2 ; Max \== Max2 ) | X::Min:Max. minimum(A,B,C):- (domain(number) -> A A=C ; B=C. maximum(A,B,C):- (domain(number) -> A B=C ; A=C. % interaction with inequalities X::Min:Max \ X ne Y <=> ground(Y), (domain(number) -> (YMax) ; (Y@Max)) | true. X::Min:Max \ Y ne X <=> ground(Y), (domain(number) -> (YMax) ; (Y@Max)) | true. X::Min1:Max \ Min2 le X <=> ground(Min2) , maximum(Min1,Min2,Min) | X::Min:Max. X::Min:Max1 \ X le Max2 <=> ground(Max2) , minimum(Max1,Max2,Max) | X::Min:Max. X::Min1:Max \ Min2 lt X <=> ground(Min2) , maximum(Min1,Min2,Min) | X::Min:Max, X ne Min. X::Min:Max1 \ X lt Max2 <=> ground(Max2) , minimum(Max1,Max2,Max) | X::Min:Max, X ne Max. % propagation of bounds X le Y, Y::Min:Max ==> var(X) | X le Max. X le Y, X::Min:Max ==> var(Y) | Min le Y. X lt Y, Y::Min:Max ==> var(X) | X lt Max. X lt Y, X::Min:Max ==> var(Y) | Min lt Y. % MULT/3 EXAMPLE EXTENSION ================================================== % mult(X,Y,C) - integer X multiplied by integer Y gives the integer constant C. constraints mult/3. mult(X,Y,C) <=> ground(X) | (X=:=0 -> C=:=0 ; 0=:=C mod X, Y is C//X). mult(Y,X,C) <=> ground(X) | (X=:=0 -> C=:=0 ; 0=:=C mod X, Y is C//X). mult(X,Y,C), X::MinX:MaxX ==> %(Dom=MinX:MaxX -> true ; Dom=[MinX|L],last(L,MaxX)), MinY is (C-1)//MaxX+1, MaxY is C//MinX, Y::MinY:MaxY. mult(Y,X,C), X::MinX:MaxX ==> %(Dom=MinX:MaxX -> true ; Dom=[MinX|L],last(L,MaxX)), MinY is (C-1)//MaxX+1, MaxY is C//MinX, Y::MinY:MaxY. /* :- mult(X,Y,156),[X,Y]::2:156,X le Y. X = X_g307 Y = Y_g331 Constraints: (1) mult(X_g307, Y_g331, 156) (7) Y_g331 :: 2 : 78 (8) X_g307 :: 2 : 78 (10) X_g307 le Y_g331 yes. :- mult(X,Y,156),[X,Y]::2:156,X le Y,labeling. X = 12 Y = 13 More? (;) X = 6 Y = 26 More? (;) X = 4 Y = 39 More? (;) X = 2 Y = 78 More? (;) X = 3 Y = 52 More? (;) no (more) solution. */ % CHIP ELEMENT/3 ============================================================ % translated to "pair domains", a very powerful extension of usual domains % this version does not work with arithmetic expressions! element(I,VL,V):- length(VL,N),interval(1,N,IL),gen_pair(IL,VL,BL), I-V::BL. gen_pair([],[],[]). gen_pair([A|L1],[B|L2],[A-B|L3]):- gen_pair(L1,L2,L3). % special cases I-I::L <=> setof(X,member(X-X,L),L1), I::L1. I-V::L <=> ground(I) | setof(X,member(I-X,L),L1), V::L1. I-V::L <=> ground(V) | setof(X,member(X-V,L),L1), I::L1. % intersections X::[A|L1], X-Y::L2 <=> intersect(I::[A|L1],I-V::L2,I-V::L3), length(L2,N2),length(L3,N3),N2>N3 | X-Y::L3. Y::[A|L1], X-Y::L2 <=> intersect(V::[A|L1],I-V::L2,I-V::L3), length(L2,N2),length(L3,N3),N2>N3 | X-Y::L3. X-Y::L1, Y-X::L2 <=> intersect(I-V::L1,V-I::L2,I-V::L3) | X-Y::L3. X-Y::L1, X-Y::L2 <=> intersect(I-V::L1,I-V::L2,I-V::L3) | X-Y::L3 pragma already_in_heads. intersect(A::L1,B::L2,C::L3):- setof(C,A^B^(member(A,L1),member(B,L2)),L3). % inequalties with two common variables Y lt X, X-Y::L <=> A=R-S,setof(A,(member(A,L),R@< S),L1) | X-Y::L1. X lt Y, X-Y::L <=> A=R-S,setof(A,(member(A,L),S@< R),L1) | X-Y::L1. Y le X, X-Y::L <=> A=R-S,setof(A,(member(A,L),R@= A=R-S,setof(A,(member(A,L),S@= A=R-S,setof(A,(member(A,L),R\==S),L1) | X-Y::L1. X ne Y, X-Y::L <=> A=R-S,setof(A,(member(A,L),S\==R),L1) | X-Y::L1. % propagation between paired domains (path-consistency) % X-Y::L1, Y-Z::L2 ==> intersect(A-B::L1,B-C::L2,A-C::L), X-Z::L. % X-Y::L1, Z-Y::L2 ==> intersect(A-B::L1,C-B::L2,A-C::L), X-Z::L. % X-Y::L1, X-Z::L2 ==> intersect(I-V::L1,I-W::L2,V-W::L), Y-Z::L. % propagation to usual unary domains X-Y::L ==> A=R-S,setof(R,A^member(A,L),L1), X::L1, setof(S,A^member(A,L),L2), Y::L2. % ATMOST/3 =================================================================== atmost(N,List,V):-length(List,K),atmost(N,List,V,K). constraints atmost/4. atmost(N,List,V,K) <=> K= (ground(V);ground(List)) | outof(V,List). atmost(N,List,V,K) <=> K>N,ground(V),delete_ground(X,List,L1) | (X==V -> N1 is N-1 ; N1=N),K1 is K-1, atmost(N1,L1,V,K1). delete_ground(X,List,L1):- delete(X,List,L1),ground(X),!. delete( X, [X|Xs], Xs). delete( Y, [X|Xs], [X|Xt]) :- delete( Y, Xs, Xt). % ALLDISTINCT/1 =============================================================== % uses ne/2 constraint constraints alldistinct/1. alldistinct([]) <=> true. alldistinct([X]) <=> true. alldistinct([X,Y]) <=> X ne Y. alldistinct([A|L]) <=> delete_ground(X,[A|L],L1) | outof(X,L1),alldistinct(L1). alldistinct([]). alldistinct([X|L]):- outof(X,L), alldistinct(L). outof(X,[]). outof(X,[Y|L]):- X ne Y, outof(X,L). constraints alldistinct1/2. alldistinct1(R,[]) <=> true. alldistinct1(R,[X]), X::[A|L] <=> ground(R) | remove_list(R,[A|L],T), X::T. alldistinct1(R,[X]) <=> (ground(R);ground(X)) | outof(X,R). alldistinct1(R,[A|L]) <=> ground(R),delete_ground(X,[A|L],L1) | (member(X,R) -> fail ; alldistinct1([X|R],L1)). % CIRCUIT/1 ================================================================= % constraints circuit1/1, circuit/1. % uses list domains and ne/2 % lazy version circuit1(L):-length(L,N),N>1,circuit1(N,L). circuit1(2,[2,1]). circuit1(N,L):- N>2, interval(1,N,D), T=..[f|L], domains1(1,D,L), alldistinct1([],L), no_subtours(N,1,T,[]). domains1(N,D,[]). domains1(N,D,[X|L]):- remove(N,D,DX), X::DX, N1 is N+1, domains1(N1,D,L). no_subtours(0,N,L,R):- !. no_subtours(K,N,L,R):- outof(N,R), (var(N) -> freeze(N,no_subtours1(K,N,L,R)) ; no_subtours1(K,N,L,R)). % no_subtours(K,N,T,R) \ no_subtours(K1,N,T,_) <=> K0,K1 is K-1,arg(N,L,A),no_subtours(K1,A,L,[N|R]). % eager version circuit(L):- length(L,N),N>1,circuit(N,L). circuit(2,[2,1]). %circuit(3,[2,3,1]). %circuit(3,[3,1,2]). circuit(N,L):- N>2, interval(1,N,D), T=..[f|L], N1 is N-1, domains(1,D,L,T,N1), alldistinct(L). domains(N,D,[],T,K). domains(N,D,[X|L],T,K):- remove(N,D,DX), X::DX, N1 is N+1, no_subtours(K,N,T,[]), % unfolded %no_subtours1(K,X,T,[N]), domains(N1,D,L,T,K). % remove*/3 auxiliary predicates ============================================= remove(A,B,C):- delete(A,B,C) -> true ; B=C. remove_list(_,[],T):- !, T=[]. remove_list([],S,T):- S=T. remove_list([X|R],[Y|S],T):- remove(X,[Y|S],S1),remove_list(R,S1,T). remove_lower(_,[],L1):- !, L1=[]. remove_lower(Min,[X|L],L1):- X@Max, !, remove_higher(Max,L,L1). remove_higher(Max,[X|L],[X|L1]):- remove_higher(Max,L,L1). % end of handler domain.chr ================================================= % ===========================================================================