[e-lang] E-Native Methods and arguments

[Kevin Reid]

On Aug 30, 2006, at 20:57, rick at vecx.net wrote:

> I have a question relating to the MethodN classes in e-native. Is  
> it important that we tie the class/functions definition to the  
> argument count? The simple question being how many arguments are  
> enough?

I am not familar with the ENative design but it sounds like you might  
not know this: in E, the number of arguments is dispatched on  
visibly, not as an implementation detail, and there are rarely  
optional/variable arguments.

? def o {
 >   to m() { println(`a`) }
 >   to m(x) { println(`b: $x`) }
 > }
# value: <o>

? o.m()
a
? o.m(1)
b: 1

Conceptually, however, this is merely a behavior of ordinary objects,  
and is not inherently required by "ELib"'s message passing system.  
For example, TextWriters have the variable-argument method 'print',  
and matchers in E-language objects can implement any dispatch rules  
they choose.

In my E-on-Common-Lisp, to (theoretically) optimize the common case,  
the message system passes 'mangled verbs' which incorporate the  
arity: :|m/0|, :|m/1|, etc. Compiled fixed-arity calls (i.e. those in  
the E language) have their mangled-verb computed at compile time;  
dynamic-arity calls (E.call(...)) compute it within E.call. The  
mangled verb is passed around as a unit unless and until it  
encounters a matcher, at which time it is disassembled and the arity  
discarded.


-- 
Kevin Reid                            <http://homepage.mac.com/kpreid/>