🚀 go-pugleaf

In article <3321@skye.ed.ac.uk> richard@aiai.UUCP (Richard Tobin) writes:
>I have never used setarg, but it seems you are confusing two things:
>the ability to change the value of an instantiated variable, and the
>ability to preserve the value of a variable when backtracking.

In my original post my point was that setarg/3 as implemented in
the compiler I'm working on works as follows (a bit hard to
write this in Prolog, of course):

	setarg(Index,Term,Arg) :-
		arg(Index,Term,OldArg),
		/* save "OldArg" somewhere */,
		/* destructively change argument number
		``Index'' of ``Term'' to ``Arg'' via
		some dirty internal mechanism */
	;	/* destructuvely change argument number
		``Index'' of ``Term'' to ``OldArg'' via
		some dirty internal mechanism */.

Thus I am perfectly correct to say a cut after *my* setarg/3
will lose a backtrack point (note the ;/2) and the dirty
internal stuff that changes the relevant arg *back* to its
original value is lost.

If we therefore use *my* setarg/3 information *can* be
transferred from one disjunctive branch to another by
means other than assert/retract.

I have no intention of *defending* the implementation
of such a dirty mechanism, which I realize destroys both
the *forward* and even *backward* declarative reading of
Prolog programs in which it is included. 

I hope this post saves some bandwidth in the long run.

-Kym Horsell