FEATURE implement '@', '++', '!' ,'--' (on the C level for meta_predicates/1)

[TeamSPoon]

See http://www.swi-prolog.org/pldoc/doc_for?object=section(2,%275%27,swi(%27/doc/packages/pldoc.html%27))

Currently all four bits are used up of the MA_INFO mask.. I am thinking about being lazy and using an 8 bit ASCII notation for storing the meta_predicate modes .. thus will be char* (instead)

[JanWielemaker]

Unsure. Paulo advocates that modes and meta-declarations should no co-exist in one description and only '*' should be used for non-meta-arguments. That makes a lot of sense to me. (Detailed) modes are anyway associated with determinism and sometimes types and thus, while each argument has a well defined meta-argument meaning, it does not have a well defined mode. The current situation is inherited from Quintus and also part of SICStus, Ciao and YAP (and possibly others).

[TeamSPoon]

On Sun, Apr 3, 2016 at 3:14 AM, Jan Wielemaker [email protected] wrote:
Unsure. Paulo advocates that modes and meta-declarations should no co-exist in one description and only '*' should be used for non-meta-arguments. That makes a lot of sense to me. (Detailed) modes are anyway associated with determinism and sometimes types and thus, while each argument has a well defined meta-argument meaning, it does not have a well defined mode.
​​

​I think this is very wise.. Since that deals with a situation like:

:-meta_predicate​(w_t_v(*,*,1,*,0)).

%% w_t_v(--VarIn:var, ++ReplaceAtom:atom, +WasGoal:callable, ?Arg:term, -NewGoal:callable) is det.
% 
% 

w_t_v(VarIn,ReplaceAtom,WasGoal,Arg,NewGoal):-
   subst(WasGoal,ReplaceAtom,VarIn,GoalMid),
  NewGoal = call(GoalMid,Arg).