% File : STRUCT.PL
% Author : Richard A. O'Keefe.
% Updated: 15 September 1984
% Purpose: General term hacking. See also OCCUR.PL, METUTL.PL.
:- op(950,xfy,#). % Used for disjunction
:- op(920,xfy,&). % Used for conjunction
/*
These routines view a term as a data-structure. In particular,
they handle Prolog variables in the terms as objects. This is not
entirely satisfactory. A proper separations of levels is needed.
*/
% subst(Substitution, Term, Result) applies a substitution, where
% ::= =
% | &
% | #
% The last two possibilities only make sense when the input Term is
% an equation, and the substitution is a set of solutions. The
% "conjunction" of substitutions really refers to back-substitution,
% and the order in which the substitutions are done may be crucial.
% If the substitution is ill-formed, and only then, subst will fail.
subst(Subst1 & Subst2, Old, New) :-
subst(Subst1, Old, Mid), !,
subst(Subst2, Mid, New).
subst(Subst1 # Subst2, Old, New1 # New2) :-
subst(Subst1, Old, New1), !,
subst(Subst2, Old, New2).
subst(Lhs = Rhs, Old, New) :- !,
subst(Lhs, Rhs, Old, New).
subst(true, Old, Old).
subst(Lhs, Rhs, Old, Rhs) :- % apply substitution
Old == Lhs, !.
subst(_, _, Old, Old) :- % copy unchanged
var(Old), !.
subst(Lhs, Rhs, Old, New) :- % apply to arguments
functor(Old, Functor, Arity),
functor(New, Functor, Arity),
subst(Arity, Lhs, Rhs, Old, New).
subst(0, _, _, _, _) :- !.
subst(N, Lhs, Rhs, Old, New) :-
arg(N, Old, OldArg),
subst(Lhs, Rhs, OldArg, NewArg),
arg(N, New, NewArg),
M is N-1, !,
subst(M, Lhs, Rhs, Old, New).
% occ(Subterm, Term, Times) counts the number of times that the subterm
% occurs in the term. It requires the subterm to be ground. We have to
% introduce occ/4, because occ's last argument may already be instantiated.
% It is useful to do so, because we can use accumulator arguments to make
% occ/4 and occ/5 tail-recursive. NB if you merely want to check whether
% SubTerm occurs in Term or not, it is possible to do better than this.
% See Util:Occur.Pl .
occ(SubTerm, Term, Occurrences) :-
occ(SubTerm, Term, 0, Times), !,
Occurrences = Times.
occ(SubTerm, Term, SoFar, Total) :-
Term == SubTerm, !,
Total is SoFar+1.
occ(_, Term, Total, Total) :-
var(Term), !.
occ(SubTerm, Term, SoFar, Total) :-
functor(Term, _, Arity), !,
occ(Arity, SubTerm, Term, SoFar, Total).
occ(0, _, _, Total, Total) :- !.
occ(N, SubTerm, Term, SoFar, Total) :-
arg(N, Term, Arg),
occ(SubTerm, Arg, SoFar, Accum),
M is N-1, !,
occ(M, SubTerm, Term, Accum, Total).
% The previous two predicates operate on ground arguments, and have some
% pretence of being logical (though at the next level up). The next one
% is thoroughly non-logical. Given a Term,
% variables(Term, VarList)
% returns a list whose elements are the variables occuring in Term, each
% appearing exactly once in the list. var_member_check(L, V) checks
% that the variable V is *not* a member of the list L. The original
% version of variables/2 had its second argument flagged as "?", but this
% is actually no use, because the order of elements in the list is not
% specified, and may change from implementation to implementation.
% The only application of this routine I have seen is in Lawrence's code
% for tidy_withvars. The new version of tidy uses copy_ground (next page).
% If that is the only use, this routine could be dropped.
variables(Term, VarList) :-
variables(Term, [], VarList).
variables(Term, VarList, [Term|VarList]) :-
var(Term),
var_member_check(VarList, Term), !.
variables(Term, VarList, VarList) :-
var(Term), !.
variables(Term, SoFar, VarList) :-
functor(Term, _, Arity), !,
variables(Arity, Term, SoFar, VarList).
variables(0, _, VarList, VarList) :- !.
variables(N, Term, SoFar, VarList) :-
arg(N, Term, Arg),
variables(Arg, SoFar, Accum),
M is N-1, !,
variables(M, Term, Accum, VarList).
var_member_check([], _).
var_member_check([Head|Tail], Var) :-
Var \== Head, !,
var_member_check(Tail, Var).
/* In order to handle statements and expressions which contain variables,
we have to create a copy of the given data-structure with variables
replaced by ground terms of some sort, do an ordinary tidy, then put
the variables back. Since we can use subst/3 to do this last step, a
natural choice of working structure in the first step is a substitution
$VAR(k) = Vk & ... & $VAR(0) = V0 & 9 = 9.
The rest is straight-forward. The cost of building the copy is o(E*V)
where E is the size of the original expression and V is the number of
variables it contains. The final substitution is the same order of cost.
For what it's worth, copy_ground(X,Y,_) & numbervars(X,0,_) => X == Y.
*/
copy_ground(Term, Copy, Substitution) :-
copy_ground(Term, Copy, 9=9, Substitution).
copy_ground(Term, Copy, SubstIn, SubstOut) :-
var(Term), !,
subst_member(SubstIn, Term, Copy, SubstOut).
copy_ground(Term, Copy, SubstIn, SubstOut) :-
functor(Term, Functor, Arity),
functor(Copy, Functor, Arity), !,
copy_ground(Arity, Term, Copy, SubstIn, SubstOut).
copy_ground(0, _, _, SubstIn, SubstIn) :- !.
copy_ground(N, Term, Copy, SubstIn, SubstOut) :-
arg(N, Term, TermN),
copy_ground(TermN, CopyN, SubstIn, SubstMid),
arg(N, Copy, CopyN),
M is N-1, !,
copy_ground(M, Term, Copy, SubstMid, SubstOut).
subst_member(SubstIn, Term, Copy, SubstIn) :-
subst_member(SubstIn, Term, Copy), !.
subst_member(SubstIn, Term, Copy, (Copy = Term) & SubstIn) :-
( SubstIn = (('$VAR'(M) = _) & _),
N is M+1 % M+1 variables seen
; N = 0 % SubstIn = 9=9
), !,
Copy = '$VAR'(N).
subst_member((Copy = Vrbl) & _, Term, Copy) :-
Vrbl == Term, !.
subst_member(_ & Rest, Term, Copy) :-
subst_member(Rest, Term, Copy).