A rogue '$CUT'/1 for Dogelog Player (Was: 5 Minute s Implementation of Prolog catch/3 and throw/1)

Hi,

Dogelog Player is a Prolog system for the
JavaScript, Python and Java target, mostly
written in 100% Prolog itself. So far we only
provided once_cleanup/2 to deal with resource
managment. It turns out we could secure a
call_cleanup/2 through a simple mechanism.

Drawing on a phrasing in a BIM Prolog document
from 1984, we could implement the post-N215
semantics of call_cleanup/2. Since we had
eliminated the native stack use from catch/3,
we could again apply patterns of continuation
manipulation for events in the continuation.

Bye

See also:

A rogue ‘$CUT’/1 for Dogelog Player
https://medium.com/2989/f41c71265331

Mild Shock schrieb:
> Hi,
> 
> Dogelog Player is a Prolog system for the
> JavaScript, Python and Java target, mostly
> written in 100% Prolog itself. Since it runs in
> cooperative multitasking enviroments, it features a
> ‘$YIELD’/1. We noticed that catch/3 causes some
> friction and we found a simple remedy.
> 
> The friction was caused by a switch from Prolog,
> to Host Language and to Prolog so that ‘$YIELD’/1
> had to create choice points. We eleminated both.
> The new catch/3 realization can easily scale to
> millions of nestings, since it has become
> independent of the native stack.
> 
> Bye
> 
> See also:
> 
> 5 Minutes Implementation of Prolog catch/3 and throw/1
> https://medium.com/2989/dc662b85fcb2