🚀 go-pugleaf

In article <1992Jul20.174202.12993@cs.sfu.ca>, gregory@cs.sfu.ca (Gregory Sidebottom) writes:
> In article <1992Jul20.083703.18058@prl.dec.com> vanroy@prl.dec.com (Peter Van Roy) writes:
> 
>    In article <1992Jul17.152828.3861@cs.kuleuven.ac.be>, bimbart@cs.kuleuven.ac.be (Bart Demoen) writes:
>    > 
>    > The implementation of ProLog by BIM has no "unsafe" variables
> 
>    Aquarius Prolog does not have unsafe variables either.
> 
> I am curious, what does it mean for a Prolog to not have unsafe
> variables.  Does this mean the compiler never uses the put_variable
> instruction?  Is a different instruction used in certain cases?  What
> about unsafe variables nested in structures?  Do 'safe' Prologs use
> more heap space?  Can somebody please clarify this for me?
> 
> Greg

Consider the two main stacks in a typical Prolog implementation: the local
stack (or "stack") which I will assume contains both environments and choice
points, and the global stack (or "heap") which contains all compound terms.

In Aquarius, the only kinds of pointers allowed are from the stack to the
heap and from the heap to the heap.  In other words, there are no pointers
from the stack to the stack, and therefore no unbound variables on the stack.
It is only these variables that are sometimes "unsafe", i.e. that have to be
moved to the heap (or "globalized") to avoid pointing to a deallocated piece
of stack.

Peter